Integration of Data and Publications - Alliance for Permanent Access

REPORT ON INTEGRATION OF DATA AND 

PUBLICATIONS 

October 17 th , 2011 

Susan Reilly a, * , Wouter Schallier a , Sabine Schrimpf b , Eefke Smit c 

Max Wilkinson d 

a 

LIBER – Association of European Research Libraries, Koninklijke Bibliotheek, 

National Library Of The Netherlands. Po Box 90407. 2509 Lk The Hague. The 

Netherlands 

b 

Deutsche Nationalbibliothek Informationstechnik, Adickesallee 1. D-60322 Frankfurt 

am Main. Germany 

c 

The International Association of STM Publishers, Prama House, 267 Banbury Road. 

Oxford OX2 7HT. United Kingdom 

d 

The British Library, 96 Euston Road. LONDON NW1 2DB. United Kingdom 

* Corresponding author: Susan.Reilly@KB.nl 

Abstract 

Scholarly communication is the foundation of modern research where empirical evidence 

is interpreted and communicated as published hypothesis driven research. Many 

current and recent reports highlight the impact of advancing technology on modern 

research and consequences this has on scholarly communication. As part of the ODE 

project this report sought to coalesce current though and opinions from numerous and 

diverse sources to reveal opportunities for supporting a more connected and integrated 

scholarly record. Four perspectives were considered, those of the Researcher who 

generates or reuses primary data, Publishers who provide the mechanisms to 

communicate research activities and Libraries & Data enters who maintain and preserve 

the evidence that underpins scholarly communication and the published record. This 

report finds the landscape fragmented and complex where competing interests can 

sometimes confuse and confound requirements, needs and expectations. Equally the 

report identifies clear opportunity for all stakeholders to directly enable a more joined up 

and vital scholarly record of modern research. 

This work is licensed under a Creative Commons Attribution 3.0 Unported License

Report on Integration of Data and Publications Grant Agreement no.: 261530 

TABLE OF CONTENTS 

EXECUTIVE SUMMARY ................................................................ 

................................................................ 

.................4 

0. INTRODUCTION 

................................................................ 

................................................................ 

........................................................ 

10 

1. INTEGRATION OF DATA AND PUBLICATIONS: GENERAL ................................................ 

12 

1.1. INTRODUCTION AND SUMMARY ................................................................................................... 12 

1.2. CURRENT PRACTICE – SOME NUMBERS RELATING RESEARCH DATA WITH PUBLICATIONS .......... 13 

1.3. WHY THE RELATIONSHIP BETWEEN DATA AND PUBLICATIONS IS SO IMPORTANT........................ 15 

1.4. THE DATA PUBLICATION PYRAMID ............................................................................................. 18 

2. RESEARCHER PERSPECTIVE OF DATA/PUBLISHING INTEGRATION: ............................ 21 

2.1. RESEARCHERS ARE THE SOURCE OF DATA ................................................................................... 21 

2.2. WHAT IS THE CURRENT PRACTICE ............................................................................................... 22 

2.3. CONCLUSIONS ON CURRENT PRACTICE ....................................................................................... 31 

2.4. IS THERE A NEED OR CASE FOR CHANGE? .................................................................................... 31 

2. How to control data(?) sharing and access ............................................................................ 33 

4. How to get credit ..................................................................................................................... 33 

5. Who pays for what .................................................................................................................. 34 

2.5. OPPORTUNITIES IN DATA EXCHANGE RELATING TO RESEARCHERS ............................................. 34 

3. INTEGRATION OF DATA AND PUBLICATIONS: THE PUBLISHERS’ PERSPECTIVE 

PECTIVE....... 

36 

3.1. HOW SCHOLARLY JOURNALS HANDLE THE INCREASING AMOUNT OF DATA ALONGSIDE THE 

ARTICLE ............................................................................................................................................. 36 

3.2. COMMON PRACTICE: SUPPLEMENTARY MATERIAL TO JOURNAL ARTICLES ................................ 42 

3.3. NEW LIMITS ON SUPPLEMENTAL FILES TO JOURNAL ARTICLES: RESTRICTIONS TO SUPPLEMENTS 

........................................................................................................................................................... 45 

3.4. HOW SAFE IS DATA IN SUPPLEMENTARY JOURNAL ARTICLE FILES? (OR: QUALITY AND 

PRESERVATION OF SUPPLEMENTARY JOURNAL ARTICLE FILES) ......................................................... 46 

3.5. DATA IN COMMUNITY-ENDORSED PUBLIC DATABASES, LINKED TO JOURNAL ARTICLES. ............. 49 

3.6. DATA STORAGE AS A SERVICE BY THE JOURNAL .......................................................................... 52 

3.7. ARTICLES WITH INTERACTIVE DATA ............................................................................................ 53 

3.8. SPECIAL DATA PUBLICATIONS AND DATA PAPERS ...................................................................... 54 

3.9. GAP ANALYSIS. ............................................................................................................................ 56 

3.10. FROM RAW DATA TO PROCESSED DATA TO DATA INTERPRETATIONS. ......................................... 58 

3.11. DIVERGING AND CONVERGING TRENDS. ................................................................................... 59 

3.12. OPPORTUNITIES FOR PUBLISHERS IN DATA EXCHANGE ............................................................ 60 

4. DATA CENTRE AND LIBRARY L 

PERSPECTIVE ................................................................ 

...... 62 

4.1. LIBRARIES AND DATA CENTRES AS CUSTODIANS OF DATA ........................................................... 62 

4.2. COMMON PRACTICE AND RATIONALE FOR ACTION....................................................................... 64 

4.3. IMPLICATIONS OF DATA INTEGRATION FOR LIBRARIES AND DATA CENTRES ................................ 69 

4.4. LIBRARIES AND DATA CENTRES ENGAGEMENT IN NEW SERVICES AND ALLIANCES ..................... 70 

4.5. GAPS AND DILEMMAS .................................................................................................................. 74 

4.6. OPPORTUNITIES FOR LIBRARIES AND DATA CENTRES .................................................................. 78 

5. REPORT EPILOGUE: MAPPING THE ROAD AHEAD 

AD ............................................................ 

81 

5.1. CAN LIBRARIES AND DATA CENTRES FILL THE MISSING LINK? ................................................... 81 

5.2. WHAT DOES THE DATA PUBLICATION PYRAMID MEAN FOR ROLES AND RESPONSIBILITIES ?...... 82 

Opportunities for Data Exchange (ODE) –www.ode-project.eu 2


5.3. IN DIALOGUE WITH RESEARCH LIBRARIANS: LIBER WORKSHOP 2011 ...................................... 83 

5.4. EMERGING ISSUES ...................................................................................................................... 85 

5.5. THE NEXT STEP: SURVEY TO DOCUMENT CURRENT AND PROJECT FUTURE ROLES .................. 87 



EXECUTIVE SUMMARY 

This report sets out to identify examples of integration between datasets and 

publications. Findings from existing studies carried out by PARSE.Insight, RIN, SURF 

and various recent publications are synthesized and examined in relation to three 

distinct disciplinary groups in order to identify opportunities in the integration of data. 

These groups are Researchers, Publishers and Libraries/Data centres. Opportunities 

identified for each group have been scoped against seven criteria: 

1. Availability 

2. Findability 

3. Interpretability 

4. Reusability 

5. Citability 

6. Curation 

7. Preservation 

Opportunities to improve the linking of data and publications have been identified for 

each stakeholder group and mapped against each of the criteria in tables at the end of 

this summary. 

Based on an examination of the available research and literature, incentives and 

barriers relating to data exchange are identified for each disciplinary group. 

The content of a draft of this report formed the basis of a workshop in June 2011 with 

professionals from research libraries. The workshop served to validate this opportunities 

and issues identified in this report. 

From a researcher perspective, the value of data is that of a first class research object 

which represents the basis of their research. Researchers discover and use data and 

analyses from others to formulate new and testable hypothesis before extending the 

evidence base with empirical data. The implications of first class research objects are 

that they require preservation, recognition, validation, curation and dissemination which 

then improve their availability, findability, interpretability and re-usability. 

Researchers perceive and enforce their creator right over the data, choose when and with 

whom they share it and wish to maintain this control. This need for control is based on 

perceived legal barriers and misuse, or absence of a trust network common in other 

forms of scholarly communication; it may be a mixture of both. Researchers want 

somewhere safe to put their data while maintaining control in order to avoid legal 

redress and professional misuse, but expect some central organisational structure to pay 

for these infrastructures. They recognise that many lack sufficient skills to manage their 

data appropriately, but, importantly, are enthusiastic to change this situation. 

Researchers see the benefit in joining publications with data in a more formal and 

agreed convention, but there must be a recognition and credit mechanism for this. They 

accept this joining as good professional practice and agree that data supporting 

traditional publication should be available with the publication. Technology can reduce 

the latency to joining data to publications but there is a lack of common best practice 



conventions for scholarly publications. Distilled into statements, our desk research has 

revealed five abstract researcher requirements for integrating data and publication. 

1. Researchers need somewhere to put data and make it safe for reuse 

2. Researchers need to control its sharing and access 

3. Researchers need the ability to integrate data and publication 

4. Researchers need to get credit for data as a first class research object 

5. Researchers need someone to pay for the costs of data availability and re-use 

Publishers are beginning to embrace the opportunity to integrate data with publications 

but barriers to the sustainability of this practice include the sheer volume of data, the 

huge variety of data formats and a question mark over exactly what data should be 

made available within, be made supplemental to or be linked with the publication. Also 

the quality of the data and attached metadata may not be consistent, lacking peer 

review, or is not being made transparent. 

The relationship between data and publications can be illustrated with a modified 

version of Jim Gray’s e-science pyramid which in this report is presented as the Data 

Publication Pyramid, see the Graph 1 below. As we descend the pyramid the exclusive 

relationship between data and publication diminishes. At the top, for example, the 

journal (and author/researcher) takes full responsibility for the publication including the 

aggregated data embedded in it and the way it is presented. For data published in the 

second layer, as supplementary files to articles, the link to the published Record of 

Science remains strong, but it is not always clear at what level the data is curated and 

preserved and if the criteria for discoverability and re-usability are met. At the Data 

Collections and Structured Database layer, the publication includes a citation and links 

to the data, but the data resides in and is the responsibility of a separate repository. The 

publication of data becomes collaborative. 

At the bottom layer of the pyramid, most datasets remain unpublished and hence 

unfindable and not re-usable. 

As Jim Gray already made clear, the data published now within or with publications, is 

only the tip of the data iceberg. 



Graph 1: The Data Publication Pyramid, developed on the basis of the Jim Gray pyramid, to 

express the different manifestation forms that research data can have in the publication process. 

See Chapter 1 for a full explanation. 

As more publishers respond to increasing author demand to making research data 

available they are focused on: 

1. establishing cross publisher best practice to make data available and retrievable 

in a persistent way 

2. collaboration with publicly endorsed community archives to make data and 

publications interlinkable 

3. presenting data in more sophisticated formats to increase reuse 

Libraries and data centres have overlapping and complementary roles in terms of data 

integration. Barriers to integration of data include a lack of policies to address the 

concerns of researchers when it comes to making their data available, the lack of 

uniformity in data preservation and curation strategies and practices. 

New publishing models linking data and articles require that libraries and data centres 

need to address particular concerns: 

1. preservation and persistence of data to ensure continued access to linked data 

2. making data findable and reusable though the use of metadata and integration 

into retrieval services 

3. working closely with researchers to encourage data sharing and best practice in 

data management 

In general, the need for action has been recognized in the library and data centre 

community. Noteworthy initiatives like the ones selected for description in this report 



(DataCite, PANGAEA, DRYAD, and Dataverse) illustrate this as well as how libraries 

and data centres support data integration. However, the degree of preparedness to take 

on the challenge varies between disciplines and between individual institutions. 

It is important that libraries and data centres act in conformance with the requirements 

of the research community, which they serve. They also need to be involved in the 

research process from the very beginning in order to ensure high data quality, which 

facilitates retrieval, usability, and preservation. 

An examination of the research and noteworthy initiatives highlights that opportunities 

exist across the three disciplines. These opportunities exist particularly in the areas of 

availability, findability, interpretability, and re-usability. 

Data Issue: Researchers opportunities (Chapter 2): 

Availability Researchers demand their data be treated as first class research 

objects 

Researchers loosen control over data 

Define roles of responsibility and control 

Findability Agree convention to propose to publishers regarding data citation 

Use of persistent identifiers such as DOI’s 

Ensure common metadata and citation practices 

Interpretability Recognize that data require metadata and work towards community 

best practice in metadata development 

Re-usability Be concerned about the long term ability for secondary use and 

consider or seek out responsible preservation actions 

Citability Agree a convention for data citation 

Follow metadata standards for datasets 


Curation Develop sustainable and realistic data management plans 

Collaboration with public data archives 

Preservation Develop sustainable realistic preservation plans 

Active engagement with public data archives 

Table 1: Data Opportunities for Researchers 



Data Issue: Publishers opportunities (Chapter 3): 

Availability Articles with data provide richer content and higher usage 

Impose stricter editorial policies about availability of underlying data 

which is in line with general funder’s trends 

Ensure data is stored in a safe place, preferably a public repository 

Be transparent about curation and preservation of submitted data 

Findability Ensure bi-directional links between data and publications 

Ensure common citation practices 

Interpretability Provide services around data such as viewer apps for underlying data 

from within the article or interactive graphs, tables and images 

Data Publications 

Re-usability Interactive data from within articles 

Links to the relevant datasets, not just to the database 


Citability Establish uniform data citation standards 




Curation Transparency about curation of submitted data 


Preservation Transparency about preservation of submitted data 


Table 2: Data Opportunities for Publishers 



Data Issue: Libraries and data centres opportunities (Chapter 4): 

Availability Lower barriers to researchers to make their data available. 

Integrate data sets into retrieval services. 

Findability Support of persistent identifiers. 

Engage in developing common metadescription schemas and common 

citation practices. 

Promote use of common standards and tools among researchers 

Interpretability Support crosslinks between publications and datasets. 

Provide and help researchers understand metadescriptions of 

datasets. 

Establish and maintain knowledge base about data and their context. 

Re-usability Curate and preserve datasets. 

Archive software needed for re-analysis of data. 

Be transparent about conditions under which data sets can be re-used 

(expert knowledge needed, software needed). 

Citability Engage in establishing uniform data citation standards. 

Support and promote persistent identifiers. 

Curation/ 

Preservation 

Transparency about curation of submitted data. 

Promote good data management practice. 

Collaborate with data creators 

Instruct researchers on discipline specific best practices in data 

creation (preservation formats, documentation of experiment,…) 

Table 3: Data Opportunities for Libraries and Researchers 



0. INTRODUCTION 

Science is changing. The massive volume and variety of data pouring out of publicly 

funded science are transforming the face of research. These data belong to everyone. If 

we manage these precious resources properly, we may tackle the Grand Challenges of 

our times – even as budgets become more restricted. It is easy to take for granted that 

data in the public domain will be protected and remain both available and accessible. 

Researchers, publishers, policymakers and funders – among many others – have started 

to appreciate that a robust, sustainably funded infrastructure is absolutely necessary to 

protect the hard-earned fruits of publicly funded research. 

Opportunities for Data Exchange (ODE) 1 , a project funded by the European Commission 

(FP7) with Grant Agreement number 261530, is gathering evidence to support and 

promote data sharing, re-use and preservation. ODE partners are members of the 

Alliance for Permanent Access (APA) and represent stakeholders with significant 

influence within their communities. ODE is identifying, collating, interpreting and 

delivering evidence for emerging best practice in sharing, re-using, safeguarding and 

citing data. ODE is also documenting drivers of change, and barriers to progress in this 

important area. 

The transition from science to e-Science is happening: a data deluge emerges from 

publicly-funded research facilities; a massive investment of public funds into the 

potential answer to thegrand challenges of our times. This potential can only be realised 

by adding an interoperable data sharing, re-use and preservation layer to the emerging 

eco-system of e-Infrastructures. The importance of this layer, on top of emerging 

connectivity and computational layers, has not yet been addressed coherently at ERA or 

global level. All stakeholders in the scientific process must be involved in its design this 

layer: policy makers, funders, infrastructure operators, data centres, data providers and 

users, libraries and publishers. They need evidence to base their decisions and shape the 

design of this layer. 

The ODE partners are: 

• European Organization for Nuclear Research (CERN) 

• Alliance for Permanent Access (APA) 

• CSC – IT Centre for Science 

• Helmholtz Association 

• Science and Technology Facilities Council (STFC) 

• The British Library 

• Deutsche Nationalbibliothek (DNB) 

• International Association of STM Publishers (STM) 

• Stichting LIBER Foundation) 

All of them are members of the Alliance for Permanent, which collectively represent all 

these stakeholder groups and have a significant sphere of influence within those 

communities. The project will identify, collate, interpret and deliver evidence of 

1 

www.ode-project.eu 



emerging best practices in sharing, re-using, preserving and citing data, the drivers for 

these changes and barriers impeding progress, in forms suited to each audience. ODE 

will: 

• Enable operators, funders, designers and users of national and pan-european e- 

Infrastructures to compare their vision and explore shared opportunities 

• Provide projections of potential data re-use within research and educational 

communities in and beyond the ERA, their needs and differences 

• Demonstrate and improve understanding of best practices in the design of e- 

Infrastructures leading to more coherent national policies 

• Document success stories in data sharing, visionary policies to enable data re-use, 

and the needs and opportunities for interoperability of data layers to fully enable 

e-Science 

• Make that information available in readiness for FP8 

This report sought to coalesce current though and opinions from numerous and diverse 

sources to reveal opportunities for supporting a more connected and integrated scholarly 

record. Four perspectives were considered, those of the Researcher who generates or 

reuses primary data, Publishers who provide the mechanisms to communicate research 

activities and Libraries & Data enters who maintain and preserve the evidence that 

underpins scholarly communication and the published record. 



1. INTEGRATION OF DATA AND PUBLICATIONS: GENERAL 

1.1. Introduction and summary 

The web, the cloud and computational capabilities in general provide an ever growing 

infrastructure for scholarly communication that makes it much easier for researchers to 

share their research data with others. At the same time, and often driven by the same 

factors, nearly all scientific disciplines have a computational stream , generating ever 

more research data. We seem to be at the verge of a Data Deluge as a recent EU report 

also concluded. 2 

We know from previous research, carried out for the project PARSE.Insight 3 , that 

around 60 % of researchers would like to use the research data of others. A similar 

reluctance for sharing as has been apparent in the interviews of WP3 of this project ODE 

that built on PARSE.Insight, where over 40 % of researchers state to have real problems 

in sharing their own data. This is further elaborated in Chapter 2: Researcher’s 

Perspective. In this sense, we may coin a new 60-40 rule: 60 % likes to get data from 

others but 40% have problems to give their own. 

Andrew Treloar, Director of the Australian National Data Service gave a talk on 28 

March 2011 at a JISC workshop in Birmingham on the management of research data 4 

and distinguished several basic problems for research data. In the handling of research 

data, he described in a cascading way, how research data are often: 

1. Unavailable, and if at all available: 

2. Unfindable, and if available AND findable: 

3. Uninterpretable. 

And even if all these 3 obstacles have been overcome, the research data found may still 

prove to be: 

4. Not re-usable. 

In this report, created and delivered in the context of project ODE (Opportunities for 

Data Exchange) we investigate how integration of data and publications can help solve 

these 4 obstacles. The questions we try to answer in this report are: How do research 

data enter the stage of scholarly communication and what are the present practices and 

policies? Where can we find important improvements for the accessibility and reusability 

of research data? What roles and responsibilities may we expect for different 

stakeholders in the scholarly information chain? 

2 John Wood, EU, 2010, Riding the Wave: http://cordis.europa.eu/fp7/ict/e-infrastructure/docs/hlgsdi-report.pdf 

3 Survey Report PARSE.Insight : http://www.parse-insight.eu/downloads/PARSE-Insight_D3- 

4_SurveyReport_final_hq.pdf 

4 Andrew Treloar at JSC workshops see 

http://www.jisc.ac.uk/whatwedo/programmes/mrd/rdmevents/mrdinternationalworkshop.aspx 



We have used 3 different perspectives to shed light on this: 

1. A researchers’ perspective, see Chapter 2 

2. A publishers’ perspective, see Chapter 3 

3. A libraries’ and data centres’ perspective, see Chapter 4 

The final chapter summarizes the outcomes of a workshop held with librarians in 

Barcelona on June 29, 2011 5 as part of the LIBER 2011 conference with a view to: 

4. Report Epilogue: Mapping the Road ahead?, see Chapter 5. 

In this general introduction, we summarize findings from previous research, such as 

PARSE.Insight and from newer desk research focused on the way research data are 

connected with publications. The hard data from the PARSE.Insight study were of high 

value for this project for 2 reasons: they span the responses from different stakeholder 

groups (researchers, publishers, librarians and data managers) and the survey was truly 

international and truly multidisciplinary. As the PARSE.Insight report points out, the 

representativity of the large spectrum of scientific disciplines was well covered, as well 

as the spread of respondents over different continents (with some dominance for US and 

Europe). This makes it one of the few studies that overarches local or mono-disciplinary 

inventories for current practices on the sharing of data. 

The value that this study attempts to add on the previous PARSE.Insight report by reuse 

of the data gathered there, is by means of a re-analysis of the survey responses from 

the perspective of data-sharing. We also included much unused data from 

PARSE.Insight that proved very relevant for this study. 

1.2. Current Practice – some numbers relating research data with publications 

In PARSE.Insight, the 2008 survey included a question: Where do you currently store 

your research data ? From the responses is clear that approximately 18 % of researchers 

submit data with the manuscripts of their publications, this is a slightly higher figure 

(but not much) than the number who deposit in archives, see the following diagram from 

Parse. 

5 Barcelona workshop: http://www.libereurope.eu/news/liber-annual-conference-and-firsteuropean-project-workshops 



Where do you currently store your research data ? (multiple answers possible) 

Graph 2: Source: PARSE.Insight 2 survey, held among researchers internationally, N = 1202 

researchers 

But this number is now growing fast, as the following evidence shows from a recent 

study into Medical Journals 6 . In that study it was found that for a sample of 28 journals 

in medicine, sampled from 138 high-impact medical journals, the number of articles 

carrying supplementary data files roughly doubled every 2 years and is now over 25 %. 

Compared to 2003, when 32 % of the journals provided the possibility to add web-based 

supplements to an article, this percentage had grown to 50 % in 2005, and to 64 % in 

2009. The number of articles offering online supplements, increased in that same period 

for these journals from 7% to 14% to the already mentioned 25%, respectively. There was 

also an increase in the percentage of journals for which at least 20% of articles have 

online-only supplements (4% in 2003 to 11% in 2005 to 32% in 2009). 

A marked increase in the number of video supplements was observed, while the largest 

share of the supplementary files is for data represented in supplementary graphs and 

tables. The number of articles with supplementary tables doubled every 2 years (10 to 22 

to 55 to 100 from 2003 to 2009), as did the number of supplementary tables (29 to 57 to 

149 to 317, the last two numbers referring to 2007 and 2009, respectively). 

From the PARSE.Insight2 survey, we also know that > 90 % of journals accepts 

supplementary material . The graph here below presents the responses from different 

publishers (small versus large). If the number of journals is factored in, it appears that 

>90 % accepts supplementary files containing research data. 

6 See as source: http://www.annemergmed.com/article/S0196-0644(10)01648-3/fulltext#sec3 



Can authors submit their underlying digital research data with their publication to you? 

Graph 3, source PARSE.Insight 2 , N= 134 publishers 

1.3. Why the relationship between data and publications is so important 

Whereas the data presented in the study quoted in the previous paragraph5 show the 

substantial amount of supplementary data added to journal articles, as well as its high 

growth rates, there is every reason to expect even more. The survey of PARSE.Insight 2 

enquired where researchers would like to submit their research data and the responses 

gave the following picture (graph 4): 

Next to the most popular category of Data Archives (of their organisation: 81 %, in their 

subject field: 60 %), more than half of the researchers (51 %) would like to submit their 

data to publishers. Of particular interest here is the marked difference between the most 

popular categories in terms of desired destinations, Graph 4, when compared to the 

actual destinations, Graph 2. Archives and publishers are the most favored, but 

remarkably the least used nowadays: in current practice, institutional archives score 

below 20 % , subject archives below 10%, publishers score a little closer to but still below 

20 %. 



Where would you be willing to submit your research data? (multiple answers) 

: 

Graph 4, Source: PARSE.Insight 2 : > 50 % of researchers would like to submit their data to 

journals, N= 1202 researchers 

The reason why researchers would like to submit their research data together with an 

article, probably relates to the way they find datasets. Again from PARSE.Insight2 we 

know that 63 % of researchers like to go to the formal literature to find and discover the 

existence of data (see also Graph 5). This option is ranked second, immediately after 

colleagues as a source (73 %) and equal to the use of general search engines (63 %): 



Where do you locate and access digital research data ? (multiple answers) 

Graph 5, Source: PARSE.Insight 2 , N=1202 researchers 



1.4. The Data Publication Pyramid 

Microsoft researcher Jim Gray is an often quoted source on the way literature and data 

would become more interrelated. He foresaw ways (graph 5) in which the e-environment 

would make it so much easier to move from the literature to data and back: 

The Jim Gray Pyramid on e-science 

Literature 

Derived and Recombined 

Data 

Raw Data 

Graph 6, Source Jim Gray on e-science 7 : Publications are only the tip of the iceberg 

For the purpose of this report, we endeavor to adapt the Jim Gray Pyramid slightly and 

introduce our own so-called Data Publication Pyramid. Four years after Jim Gray 

expressed his thoughts, we see a new wave of practices emerge where the literature is in 

fact integrating literature and data or at least making its best attempts. This Data 

Publication Pyramid (graph 1) aims to show the different manifestations that data can 

undergo when published within or in the context of publications, or even when not 

published at all (but remaining in drawers and on disks of the institute): 

7 http://research.microsoft.com/enus/collaboration/fourthparadigm/4th_paradigm_book_jim_gray_transcript.pdf 



Graph 1 repeated, The Data Publication Pyramid, based on the Jim Gray Pyramid. 

We wish to emphasize at this point that this pyramid does not describe which stages of 

manifestation research data can go through in their evolution towards reusable data 

products. The main purpose of this pyramid is to explain the different manifestations 

research data can have in the context of their availability within, with, supplementary to 

or referenced from an official scholarly article as the main manifestation of the record of 

science. 

In Chapter 3 (Integration of Data and Publications), in Chapter 4 (Data Centre and 

Library perspective) and Chapter 5 (Next Steps, the Road ahead) this Data Publication 

Pyramid is explained further and used to distinguish different best practices as they 

currently occur. 

In short, we can summarize the benefits of good integration of articles and research data 

as follows, along the key issues raised by Andrew Treloar3 and tabulate obstacles and 

the solutions provided by integrating Data and Publication (Table 4) 

• publications help the data to be better discoverable 

• publications help the data to be better interpretable 

• publications provide the author better credits for the data 

• and reversely: the data add depth to the article and facilitate better 

understanding. 



Obstacle 

Data are unavailable 

Data are unfindable 

Data are uninterpretable 

Data are not re-usable 

Integration of Data and Publication helps to 

Indicate availability of data 

Links from the publication help locate datasets 

Publications will describe and explain datasets, links 

from data to publications help interpretability of 

datasets 

Description in article can improve usability or the 

article can provide access to a re-usable set, perhaps 

even offer an API to datasets ? 

Table 4: Data Obstacles 

As will be explained in the following chapters, there are 3 additional necessary 

conditions that need to be fulfilled for good accessibility and re-usability of research 

data. They are: 

1. Citability 

2. Curation 

3. Preservation 

These elements are also addressed in-depth in the following chapters and used as 

criteria for drivers and potential opportunities for different players in the scholarly 

communication landscape. Included in the following chapters are case studies for 

laudable initiatives, reported unsolved issues, and opinions and desires as expressed to 

the project team in email exchanges and interviews from the perspective of researchers, 

publishers and libraries/data centres. 



2. RESEARCHER PERSPECTIVE OF DATA/PUBLISHING INTEGRATION: 

2.1. Researchers are the source of data 

Researchers are able to generate more data than ever before. This has partially been 

driven by technological advances that increase the accuracy, sensitivity and multiplicity 

of empirical data collection across disciplines8. But equally, researchers have been able 

to track, aggregate, abstract, transform and generally re-purpose existing data to drive 

forward data driven research. Facilities and infrastructures that increase our 

understanding of sub-atomic particles or extend our reach into the universe have lead to 

way in generating data; they have been quickly followed by social, environmental and 

biomedical disciplines that capture complex and uncertain data for modelling biological 

processes, social dynamics and environmental forecasting. Data collected in vivo or in 

situ and modelled in silico reveal promising new ways in understanding and intervening 

in all manner of human behaviour for benefit9’10. All these data are a fundamental 

component of scholarly communication and are the evidence that underpin scholarly 

publication. Their value extends beyond original use and many represent substantial, 

non-reproducible and valuable intellectual assets to many stakeholders1112. Our ability 

to manage and maintain such digital data have not always kept pace with our ability to 

generate it and presents a risk to modern research; we are in danger of losing the ability 

to link the evidence base that support scholarly publication and as a consequence break 

the cycle of scholarly communication. 

From a researchers perspective the value of data is that of a first class research object 

which represents the basis of their research. Researchers discover and use data from 

others and analyse them to formulate new and testable hypothesis before extending the 

evidence base with empirical data. The implications of first class research objects are 

that they require preservation, recognition, validation, curation and dissemination; in 

doing so they become more available, discoverable, interpretable, re-usable and citable. 

This section of our report will review contemporary evidence of current practice from a 

researcher’s perspective and investigate any need for change. From these needs we will 

8 

Hanson B, Sugden A, Alberts B (2011). Making data maximally available. Science. 

331(6018):649 

9 Editorial: Crowdsourcing human mutations. Nature Genetics 2011, 43(4):279 

10 Giardine B, Borg J, Higgs DR, Peterson KR, Philipsen S, Maglott D, Singleton BK, Anstee DJ, 

Basak AN, Clark B, Costa FC, Faustino P, Fedosyuk H, Felice AE, Francina A, Galanello R, 

Gallivan MV, Georgitsi M, Gibbons RJ, Giordano PC, Harteveld CL, Hoyer JD, Jarvis M, Joly P, 

Kanavakis E, Kollia P, Menzel S, Miller W, Moradkhani K, Old J, Papachatzopoulou A, 

Papadakis MN, Papadopoulos P, Pavlovic S, et al.(2011). Systematic documentation and analysis 

of human genetic variation in hemoglobinopathies using the microattribution approach.Nature 

Genetics 43(4):295-301 

11 Mons B, van Haagen H, Chichester C, Hoen PB, den Dunnen JT, van Ommen G, van Mulligen 

E, Singh B, Hooft R, Roos M, Hammond J, Kiesel B, Giardine B, Velterop J, Groth P, Schultes E, 

(2011). The value of data. Nature Genetics 43(4):281-3 

12 Curry A (2011). Rescue of old data offers lesson for particle physicists. Science. 

331(6018):694-5 



seek opportunities in the form of enablers, drivers and incentives to support such 

change. 

2.2. What is the current practice 

Data has always been part of scholarly communication though increasingly scholarly 

publication has been unable to maintain the join to this evidence base 13 . Integrating 

data and publications implies an outcome of data sharing. Thus a researcher perspective 

is primarily concerned with data sharing; the integration of scholarly communication is 

essentially an enabler of this. In order to understand why data and publications risk 

moving apart it is essential to understand what causes such divergence and whether 

these are barriers we can influence. 

Illustrated in Graph 4 in the Introduction Chapter, the PARSE.Insight 14 survey provided 

compelling and insightful evidence regarding the perspective of researchers on current 

practice in data management and sharing. With support from additional and 

contemporary reports ,15 , and publications a strong picture of researcher attitudes to data 

management and requirements begins to emerge. 

Researchers find data in the same way as other information (via literature, 

colleagues) and IT has added to these rather than supplanted any one common 

practice activity. 

The methods used to disseminate and validate research seems tied to common 

research practices where colleagues and formal literature predominate. Search 

engines and institutional databases suggest that technology has made this easier 

than their analogue counterparts of catalogues and libraries. This is in 

agreement with the report of the UK’s Research Information Network (RIN) in 

which technology was used to increase efficiency in behaviour, in the life sciences, 

rather than supplant it 16 . Data archives offer a significant source of discovery but 

it is difficult to determine if this was in addition to, or independent of the more 

common discovery through literature and colleagues. 

13 

Aalbersberg, IJ, Kähler, O. (2011)Supporting Science through the Interoperability of Data and 

Articles , D-Lib. Volume 17, Number ½ doi:10.1045/january2011-aalbersberg 

14 

http://www.parse-insight.eu/index.php 

15 

If you build it they will come. How Researchers perceive and use web2.0. A Reseatrch 

Information Report, July 2010. 

16 

Patterns of information use and exchange. Case studies of researchers in the life sciences. 

Research Information Network, November 2009. 



Researchers want control over the extent and manner of sharing their data, 

consistent with the concept of data as a professional asset. 

This particular finding of data largely being shared with colleagues but only for 

25 % of the respondents sharing it openly with everyone and more than 20 % not 

sharing any, re-enforces the notion that researchers want to control their data 

How openly available is your data? 

My data is openly available for my research group / 

colleagues in research collaboration. 

58% 

My data is openly available for everyone. 

25% 

Access to my data is temporarily restricted. 

I do not share my data, but I would like to do so in the future. 

My data could be made available with appropriate changes 

(e.g. anonymous clinical data) 

My data is openly available for my research discipline. 

16% 

16% 

11% 

11% 

I do not share my data and I do not want to share it in the 

future. 

My data is available for a fee. 

4% 

6% 

and reject the concept of un-controlled open data sharing. 

0% 10% 20% 30% 40% 50% 60% 70% 

Graph 7: Which of the following applies to the digital research data of your current research 

N=1270 (PARSE.Insight 14 ) 

They are open with their nearest colleagues and when their professional practice 

mandates it (e.g. publisher policy or good practice), though the extent of data sharing 

decreases rapidly as their potential control over the data decreases, i.e. they wish to 

choose when and with whom they share their data (Graph 7). 

However the researchers’ practice becomes more complex when questioned about their 

specific data needs, i.e. not what they do with their own data but rather what they 

require from others. 

According to the responses to the PARSE.Insight survey, 63% of the researchers would 

like to make use of data gathered by other researchers in their discipline (N=430) while 

70% of respondents already do (N=638). Interestingly, when asked about data gathered 

by researchers in other disciplines, still 40% would like to make use of it (N=689 ), and 

46% state that they do so already (N=1264). 



Some researchers have identified a requirement in skills training or resources to assist 

in the safe and appropriate maintenance processes of data management 17 and this is 

likely to be reflected across other disciplines. Several organisations are producing guides 

and tools to assist with this, but it is unclear if they are making as big an impact as they 

perceive. A key to this was the extent and use of data standards together with their 

effective uptake profile. A great majority of respondents claimed to have used no 

standards when offered a selection of those in current use (Graph 8). 

Knowledge of standards and guidelines 

Other 

3% 

OAIS 

6% 

Dublin Core 

7% 

PREMIS 

3% 

OAI-PMH 

3% 

OAI-ORE 

3% 

METS 

3% 

MPEG21-DIDL 

3% 

None 

67% NISO 

2% 

Graph 8: Which of the following standards or guidelines that are used in digital preservation are 

you familiar with? n=1202(PARSE.Insight 14 ) 

It is well known that the development of community standards is a slow and demanding 

process. Agreement is based on function and often there are many opinions on 

pragmatic utility than can be accommodated in an easily implemented standard. There 

may be a case for supporting standards development as a community activity more 

actively, though how this could be achieved needs to be defined. 

17 

Reichman OJ, Jones MB, Schildhauer MP (2011). Challenges and opportunities of open data in 

ecology. Science. 331(6018):703-5 



Researchers perceive legal and professional reasons for not sharing their data 

Barriers for sharing research data 

Legal issues 

Misuse of data 

41% 

41% 

Incompatible data types 

33% 

Lack of technical infrastrcuture 

Lack of financial resources 

fear to Lose scientific edge 

28% 

27% 

27% 

Restricted access to data archive 

21% 

No problems foreseen 

16% 

Other 

10% 

0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 

Graph 9: source PARSE.Insight ; Do you experience or foresee any of the following problems in 

sharing you data? N=1270(PARSE.Insight 14 ) 

Barriers to sharing data, whether real or perceived, were mixed between reasons for not 

sharing and reasons for not using certain data types or particular data sources. For 

example, Graph 9 suggests many respondents suggested there were legal barriers to 

their sharing, though it was unclear if these were fear of prosecution or responsibility for 

IPR particularly in biomedical sciences that involve human subjects or commercial 

potential 18 . Licensing research data is a recognised as a complex and time consuming 

activity 19 and there is a need to simplify and streamline the process by which 

researchers or those in control of research data assert control over their data assets, an 

opinion supported by the recent Hargreaves report in the UK 20 . The next most frequent 

barrier was a fear of misuse, which may include a validation threat to analyses that 

contradict the original findings, discovery of additional findings or exposure of the data 

creator to legal redress, thus strongly associated with the most common response of legal 

issues. For example repurposing data that leads to breaching data protection legislation 

could hold the data publisher or those responsible for the data vulnerable to criminal 

proceedings. Finally, incompatibilities between data and lack of a financial and 

technical infrastructure were cited as strong barriers to sharing, long recognised as a 

18 

Mathews DJ, Graff GD, Saha K, Winickoff DE (2011) Access to stem cells and data: persons, 

property rights, and scientific progress. Science. 331(6018):725-7 

19 

Alex Ball (DCC) 2011. How To Licence Research Data. A Digital Curation Centre and JISC 

Legal ‘working level’ guide. 

20 

Digital Opportunity A Review of Intellectual Property and Growth, An Independent Report by 

Professor Ian Hargreaves May 2011 



result of information technology and computational ability moving faster that our 

capability for data management and planning. 

Taken together these results indicate researchers are willing to use others data 

providing they can validate them, but are wary about openly sharing their own. This 

need not be seen as a closed relationship. It could well indicate deeper levels of sharing 

enablers than simply altruistic motivation, e.g. attribution, provenance and reliability. 

Researchers want credit 

Even if data can be shared or published there was, as expected, almost universal 

recognition that a ‘credit to the data creator’ facility must exist (Graph 10). Good 

research practice requires recognition for intellectual contributions and these should 

include data. In the same way citation of traditional publications play for recognising 

individual intellectual work, a similar convention is required for data though no agreed 

convention exists 21,22 . 

Do you want to be credited when your underlying digital research data is used by others? 

(PARSE.Insight) 

No 

8% 

Yes 

92% 

Graph 10: Source PARSE.Insight 14 , N=1171 

However, when it comes to clear guidelines about how data can be made citable and how 

this can be integrated with traditional publications, there are few examples of common 

standards or shared good practice. This is supported by the findings in PARSE.Insight, 

where the vast majority of researchers understand the benefits of a joined scholarly 

communication but were are unaware of publisher policies and these policies (see 

Graphs 11 & 12 and Chapter 3). 

21 

Starr,J. Gastl, A (2011) isCitedBy: A Metadata Scheme for DataCite. D-Lib. Volume 17, 

Number ½ doi:10.1045/january2011-starr 

22 

Brase, J and Farquhar, A (2011). Access to Research Data . D-Lib. Volume 17, Number ½ 

doi:10.1045/january2011-brase 



Do journals to which you typically submit t your work require you to include relevant 

digital research data (i.e. data used to create tables, figures, etc.)? 

n=129(PARSE.Insight 14 ) 

Yes 

19% 

No 

81% 


Do you think it is useful to link underlying research data with formal literature?n- 

2289(PARSE.Insight14 

14) 

Link underlying research to formal literature. 

No 

15% 

Yes 

85% 


Many researchers see the problems with data getting worse 

The problem of data volume is illustrated by a steady increase in expectancy from 

research together with an increasing ‘don’t know’ cohort in the PARSE.Insight 

investigations of data volume expectations. The figures for data volumes below 1Gb-1Tb 

exhibit decreasing expectancy while those about 1Gb-1Tb are increasing expectancy 

(Graph 13). 



45% 

40% 

35% 

30% 

25% 

20% 

15% 

10% 

5% 

0% 

Estimated amount of data stored per research project 

40% 41% 

36% 

25% 

17% 

19% 

20% 

13% 

13% 

8% 

5% 

6% 

3% 

5% 

1% 1% 2% 

1% 

2% 

0% 0% 

17% 

14% 

11% 

0MB 1-100MB 100MB-1GB 1GB-1TB 1TB-1PB 1PB-10PB >10PB Don't Know 

Current In 2 years In 5 Years 


Together with data volume expectancy there are well understood data preservation risks 

that include lack of infrastructure and custodian roles, indicated by data preservation 

issues being perceived as either important or very important by the majority of 

respondents (Graph14). 



Threats to digital preservation 

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 

Lack of sustainable hardware, software or support of computer 

environment may make the information inaccessible 

41 

39 

13 

5 1 

The current custodian of the data, whether an organisation or 

project, may cease to exist at some point in the future 

36 

42 

15 

5 

3 

Users may be unable to understand or use the data e.g. the 

semantics, format or algorithms involved 

34 

42 

16 

6 

3 

Evidence may be lost because the origin and authenticity of the data 

may be uncertain 

33 

44 

17 

4 2 

Loss of ability to identify the location of data 

25 

44 

23 

5 

4 

The ones we trust to look after the digital holdings may let us down 

20 

37 

26 

10 

7 

Access and use restrictions (e.g. Digital Rights Management) may not 

be respected in the future 

19 

37 

25 

13 

6 

Very important Important Slightly Important Not important Don't Know 

Researcher perceived threats to digital preservation(n=1201-1210) 1210) (PARSE.Insight 14 

Graph 14: Source PARSE.Insight 14 , N=1201-1210 

14 ) 



Researchers s believe others should pay 

There are strong views, possibly reinforced by an appreciation of preservation and 

hardware considerations, that once the data have been generated and used, their 

preservation and archiving responsibilities should rest with other organisational 

structures, with the exception of a non-specific ‘research community’ being identified. 

These were firm views as few respondents either didn’t know or selected ‘other’ (Graph 

15 &16). 

Who should pay for preservation of publications? 

National government 

57% 

Commercial organisation 

42% 

Research community 

European Union 

My organisation 

34% 

32% 

35% 

Don't know 

Other 

5% 

6% 

0% 10% 20% 30% 40% 50% 60% 

Graph 15: Source PARSE.Insight 14 , N=1188: Who, in your opinion, should pay for the preservation 

of publications?, multiple answers possible) 

Who should pay for preservation of digital research data? 

National government 

61% 

My organisation 

41% 

European Union 

36% 

Research community 

28% 

Don't know 

15% 

Commercial organisation 

10% 

Other 

4% 

0% 10% 20% 30% 40% 50% 60% 70% 

Graph 16: Source PARSE.Insight 14 , N=1188: Who, in your opinion, should pay for preservation of 

digital research data? (multiple answers possible) 



2.3. Conclusions on current practice 

Taken together there appears to be a complex relationship between the researcher and 

the data they collect and create. Researchers perceive and enforce their creator right 

over the data, choose when and with whom they share it and wish to maintain this 

control. This need for control appears based on perceived legal barriers and risk of 

misuse, or absence of a trust network common in other forms of scholarly 

communication; it may be a mixture of both. Researchers want somewhere safe to put 

their data while maintaining control in order to avoid legal redress and claims of misuse, 

but expect some central organisational structure to pay for these infrastructures. In a 

study of data sharing in the biomedical informatics domain, a training review indicated 

that many researchers recognise they lack sufficient skills to manage their data 

appropriately, but importantly are enthusiastic to change this situation23. Researchers 

would benefit in joining the publication with the data in a more formal and agreed 

convention, and recognition and credit mechanism for this can help as important drivers 

and incentives. They accept joining data to publication as good professional practice (see 

graph above) and agree that data supporting traditional publication should be available 

with the publication. Technology can reduce the latency to joining data to publications, 

but the policies of the publishers requiring the availability of data supporting 

publications are so far very much in a pioneering stage (see chapter 3). 

2.4. Is there a need or case for change? 

Data and publications belong together24 and researchers are the link between these two 

established intellectual objects. The evidence that supports scholarly discourse cannot 

be lost without severe consequences for scholarly communication. Distilled into 

statements, our desk research has revealed five abstract researcher requirements for 

integrating data and publication. 

1. Researchers need somewhere to put data and make it safe for reuse 

2. Researchers need to control its sharing and access 

3. Researchers need the ability to integrate data and publication 

4. Researchers need to get credit for data as a first class research object 

5. Researchers need someone to pay for the costs of data availability for re-use 

1. Where to put data and make it safe for reuse 

Research data centres exist but they are fragmented and operate in different ways. 

Generally data centres are community or discipline focussed and where significant 

investment is available more large scale operations are established (as is evident in 

particle physics and astronomical disciplines). Such mature data archives are capable of 

taking full responsibility for the data they hold with clear preservation and access 

23 http://www.cancerinformatics.org.uk/training.html 

24 Smit, E (2011), Abelard and Héloise: Why Data and Publications Belong Together D-Lib. 

Volume 17, Number ½ doi:10.1045/january2011-smit 



policies, with some making use of community developed quality tools like the Data Seal 

of Approval 25 . These organisations can either be concerned with ‘big data’ in the case of 

particle physics and astronomical data centres, or complex social data, as in the case of 

UK Data Archive. In contrast there are numerous ad hoc collections of community 

activities borne out of immediate need and uncertain future. These so-called ‘long tail’ 

data centres in numerous and low level data generating disciplines, e.g. ecology, 

evolution etc have the potential to produce more volume and more complex data for 

which ‘big data’ solutions will not be appropriate. In turn these ‘long tail’ data centres 

will likely require much more resource to both establish and maintain. 

Sustainability models for various data needs across disciplines would assist in 

determining how much resource is required for what type of data and for how long. A 

number of pilot and low scale projects have attempted to establish this alongside the 

well known and established initiatives for ‘big’ data 26 

In summary, the research data landscape is both large and complex. PARSE.Insight 

provides evidence that these examples are not enough: researchers are either unaware 

(54 %) or know there is no facility (37 %) to put and maintain their data. 

Is there a preservation facility for preserving digital research data 

which can be used by all projects 

Yes 

within your discipline? 

9% 

No 

37% 

Don't Know 

54% 

Graph 17: Source PARSE.Insight 14 , N=1198 Is there a preservation facility for preserving digital 

research data which can be used by all projects within your discipline?) 

25 Klump, J.(2011). Criteria for the Trustworthiness of Data Centres D-Lib. Volume 17, Number 

½ doi:10.1045/january2011-klump 

26 www.datadryad.org 



2. How to control data(?) sharing and access 

Disciplines vary widely; the RIN report on patterns of information use and exchange 

provided compelling evidence from in depth interviews on the differences between 

sharing practices across life science disciplines 16 . There appeared many levels of control 

that were either encountered (while looking for data) or imposed (when the custodian of 

data). What this suggests is the perception of and need for data sharing are confounded 

by confusion over the risks of data sharing. With the PARSE.Insight data suggesting a 

fear of legal redress and misuse of data as the main concerns it would seem that 

ownership and responsibility for data have become the default stance over authority; 

without a clear accreditation or registration framework for data, like citation, or a open 

sharing environment with accountability, reluctance to share wins out over professional 

transparency. Researchers have ownership of data as a consequence of generating it, 

but transferring licence or ownership to use or share data is a confusing barrier and if 

there are any likelihoods of exposure to prosecution for either data misuse or unethical 

data collection then it is easier to simply close access to the data and severely limit any 

sharing. No clear authority or accountability role is available for research data outside 

its ownership. 

So as data enters scholarly communication the issue arises; who will be responsible for 

it, securing the authority to persist, preserve, share it and take legal responsibility. 

Some co-ordination and advice centres exist. The DCC 27 in the UK was established to 

support UK researchers engaged in digital research activities. It has since expanded 

into international partnerships and provides a rich resource for any researchers or 

institutional service (eg digital repository or similar) and offers advice on all manner of 

data curation issues including licensing data. In addition the UK Data Archive releases 

a series of best-practice guidelines for researchers from the social sciences that can be 

applied more widelyError! Bookmark not defined.. 

3. The capability to integrate data and publications 

Many researchers see the benefit of integrating the scholarly record but little best 

practice conventions exist. In contrast to the practicalities of what to do with data there 

are a growing number of examples of how to re-join data to the publications they support 

(see Chapter 3). 

4. How to get credit 

Almost all respondents in PARSE.Insight agreed that should their data be used they 

should receive credit for it. This is in line with the professional impact that researchers 

receive from publication activity. Presently there is no agreed framework for citation of 

data nor a capability to measure impact in a similar manner to the way traditional 

citation impact factors are created. Thus the question is raised, does data need a specific 

27 

http://www.dcc.ac.uk/ 



citation/recognition framework or is the current framework sufficient to absorb the 

requirements of data citation? DataCite 28 believe data are first class research objects 

with separate requirements from scholarly publication and as such require a citation 

framework that can accommodate these. DataCite is an international association that 

is implementing such a framework independent of discipline to support data citation via 

the registration of persistent identifiers (DOI’s) that enable linking to and from these 

data sets (see also Chapter 4). 

5. Who pays for what 

Researchers feel that while they recognise data preservation and archiving cost money, 

they are unable to pay for it. In fact the complex processes of data preservation are well 

understood in many disciplines outside research, especially those where severe and 

expensive legal requirements are imposed, e.g. financial institutions are required to keep 

data for many decades, nuclear installations are required to keep digital records and 

data for perhaps centuries. In the UK at least, the cost of both data preservation and 

data sharing is recognised in many policies from the Research Councils. Assisting in 

developing and promoting these policies with well documented and realistic scenarios 

and use cases for best practice data management throughout the research process would 

be advantageous to all . 

2.5. Opportunities in data exchange relating to researchers 

In summary, returning to the criteria over which we attempt to identify opportunities for 

data exchange, from a researcher’s perspective the following have been identified. 

Data Issue: 

Availability 

Findability 

Interpretability 

Re-usability 

Citability 

Researchers opportunity to help improve situation: 

Researchers demand their data be treated as first class research 

objects 

Researchers loosen control over data 

Define roles of responsibility and control 

Agree convention to propose to publishers regarding data citation 



Recognize that data require metadata and work towards community 

best practice in metadata development 

Be concerned about the long term ability for secondary use and 

consider or seek out responsible preservation actions. Further, 

consider this as part of good research practice rather than as a closing 

activity. 

Agree a convention for data citation 



28 

http://www.datacite.org 



Curation 

Preservation 

Develop sustainable and realistic data management plans 


Develop sustainable realistic preservation plans 

Active engagement with public data archives 

Table 1 repeated: Data Opportunities for Researchers 

These have been further broken down to incentives, drivers and enablers. 

Incentives 

Joining functions that complete scholarly communication by integrating data and 

publications 

Citation framework that encourages credit, attribution and re-use to remove hesitation 

on the researchers side 

Review/validation process (e.g. data journal) that support trust 

Drivers 

Impact and re-use metrics that support incentives 

Data management plans/Data sharing plans as part of start up activity 

Disentangled responsibility between data creators and data custodians 

Enablers 

Clear and consistent IPR and other rights statements from stakeholders 

Mandated infrastructures like data centres and data archives that can persist and 

preserve 

Recognition frameworks that support data as first class research objects 

Submission processes that minimise overheads and effort 

Embedded training activities and practices that focus on data management skills rather 

than simply data manipulation skills 



3. INTEGRATION OF DATA AND PUBLICATIONS: THE PUBLISHERS’ 

PERSPECTIVE 

3.1. How scholarly journals handle the increasing amount of data alongside the article 

Building on the previous chapters of this report, and well aware of the desire of 

researchers to publish their data in a citable way and to find data via the formal 

literature, we focus here on the ways publications and data are being integrated. In our 

pursuit of present practices and new initiatives in the field of STM journals, we have 

encountered (and will describe in this chapter) the 4 basic categories for different ways 

in which data and publications can be connected and/ or integrated. They follow largely 

the Data Publication Pyramid as presented in the Introduction Chapter and the first 4 

categories listed there (while category 5: data in drawers and on disks, concerns the 

category of unpublished data, which is addressed further in Chapter 5): 

Graph 1 repeated: The Data Publication Pyramid, see chapter 1 for a full explanation. 

This is a high level overview of each of these categories before they are described more 

in-depth in the remainder of the chapter, with several examples: 



1. Data contained within peer reviewed articles 

This is the traditional publishing model in which the researcher fully analyzes and 

processes the data and describes the conclusions derived from them in the scholarly 

article. The conclusions drawn from the data are illustrated by summarizing the 

relevant data (or data-outcomes) in tables, graphs and other illustrations, and, 

increasingly, also in multimedia applications. 

Advantages are in the tight embedding and integration of the data into the scholarly 

record, citable and retrievable as such, available to all researchers and users. Authors 

get all credits for their article. 

Limitations are that these present a high level of aggregation of the data, data are hard 

to find separately from the article and usually not in a re-use friendly way. 

2. Data resides in supplementary files added to the journal article 

Nearly all STM journals offer authors the service to add in supplementary files to their 

article any relevant material that is too big or that will not fit the traditional article 

format or its narrative, such as large datasets, multimedia files, large tables, 

animations, high resolution files, protocols, large bibliographies, etc. With the increased 

computational nature of many disciplines, the use of supplementary files has increased 

sharply recently (see also Chapter 1, Introduction). 

Advantages of using supplementary files are that the volume of the data is no longer an 

issue and that the data are still closely tied to the official scholarly record and remain 

citable, while authors are no longer restricted by the article format. It makes optimal use 

of online facilities. 

Limitations are that file size is usually not much larger than 10 GB and that, from an 

author’s perspective, the curation and preservation of the supplementary files is not 

always clear. Few standards exist between journals on how to indicate the presence of 

supplementary files or where to find them. Only in few cases will the supplementary 

material be provided with a separate DOI or other persistent identifier to enable linking 

independent from the main article. 

a. Journal articles offer supplementary files with extra data, but restricted 

in volume and format 

A sub-category under Supplementary data files exists for those journals that have 

restricted the use of supplements for this purpose. The first examples of high standing 

journals which can no longer manage the sheer volume of materials in supplementary 

article files and as a consequence have put limitations on what can be included in 

supplementary files (Cell) or even who no longer accept supplementary material at all 

(Journal of Neuroscience) other than multimedia files that should be considered integral 

to the article content. 



Advantages are that any supplements of a journal articles remain tied to the official 

scholarly record, and are part of its peer review process, but 

Limitations are in the new restrictions posed; adding original research data is often no 

longer possible. 

3. Data resides in Community-endorsed Public Repository with bi-directional 

linking to and from articles 

In this model the data relating to a scholarly article are deposited in designated Public 

Repositories, best examples are GenBank and World Protein DataBank. The accession 

numbers of the data in those databases are added to the journal manuscript and 

referenced, often within the article text, as well as in the footnotes or reference list. 

Advantages are that the data become part of larger datasets in the same area thereby 

serving the research community as a whole, and it is normalized, standardized, curated 

and preserved. The connection between data and publication are secured via the 

accession numbers that are embedded in the article. Even better are the examples 

(Pangaea, CCDC, PubChem) where the bidirectional linking between data and articles 

are secured, and likewise from the data to the articles. There are no restrictions on 

volume. 

Limitations are that these databases tend to exist only in a few subject areas so far, 

mainly biology, life science, earth science and chemistry. The future of these databases 

often depends on government funding and may be threatened by budget cuts. 

a. Journals have set up an own storage facility for data 

A sub-category for data referenced from articles that reside in a special data-storage 

facility established by the publisher fits the example of Thieme publishers in Germany, 

as later described in paragraph 3.6. Thieme have recently instigated collaboration with 

the data facilities of FIZ Karlsruhe to offer their chemistry authors the possibility to 

store the raw and original research data alongside their articles. 

Advantages lie in the strong link between the data and the scholarly record and the 

availability of the data for further examination and re-use to all the journal readers, 

editors and other users. 

Limitations are the danger of creating new silos for data per journal or per publisher, 

which can be a barrier to discoverability and reuse. 

b. Journals offer dynamic data made interactive, data can reside with the 

article or in public repositories 

Another sub-category exists for journals who present the relevant data sitting in an 

official repository or data centre or anywhere else from within the article. This model 

emphasizes what readers of a journal article can do with the data rather than where the 



data actually resides. Via click throughs from graphs and tables, readers can play with 

the underlying data and their visualizations. As an example, the BioChemical Journal 

from Portland Press does this via dynamic pdf’s. Elsevier has some examples of data 

viewers that work from within the article but using data in Genbank and PDB bank. 

This may increasingly emerge as a model for Linked Open Data and the emerging data 

web. 

Advantages are clearly in reuse and in increased interpretability of underlying data. 

Data become re-usable within the context of the scholarly record. 

Limitations are in the open availability of these data. Applications are usually only used 

for the data within the article, i.e. category 1 of this list. 

4. Journals dedicated to so-called Data Publications only 

In this model the journal publishes descriptive articles about datasets that are usually 

stored in a repository. The description of the data generation and its potential use allows 

the authors credits for their work while strongly promoting the interpretability and 

re/use. Examples are the Earth Science Systems Data Journal (ESSD) and the newly 

launched journal Gigascience. Other, already existing journals offer hybrid models in 

which they have opened up for descriptive data articles (Int Journal of Robotics) as a 

new article type next to the traditional research papers. 

Advantages are in the credits for the author, the citability and the reuse. 

Limitations are in the challenges for high quality peer review, very few peer review 

standards exist so far for datasets and their descriptions. The system depends on proper 

and persistent bi/directional linking. 

We can summarize and compare these 4 categories (plus 3 sub-categories) in the 

following table. Subsequent paragraphs contain more extensive descriptions of current 

practices and policies. 



Data (selections) underlying 

articles reside in: 

Data contained within the peer 

reviewed article, in tables, 

graphs, plotting, etc 

Supplementary files to journal 

articles – whatever it contains 

with very few restrictions on size 

and format 

2.a Supplementary files to 

journal articles, with restrictions 

and tightened instructions to 

manage the proliferation of 

supplemented material 

Advantages: 

Analyzed data and relevant data 

selections are integral part of the Record 

of Science. 

Readers, users and peer reviewers can 

find and consult these data selections 

Datasets and publications tightly 

connected, data is embedded in public 

record of science, managed and preserved 

as such, author gets full credits, reviewers 

and readers are able to access data in 

combination with the article 

More clarity on the supplemental 

materials that journals can and will 

support. Better reassurance of curation 

and preservation and perpetual access. 

Journals will encourage authors to place 

unsupported materials in a reliable 

repository 

Limitations: 

Usually high level of aggregation of the 

data, more data summaries than full 

set of original data. 

Usually not findable or retrievable 

separately from the article. 

Not well reusable outside the context of 

the article. 

Volume is a limitation, usually datasets 

not bigger than 10MB. 

Curation sometimes unclear, 

preservation likely to remain limited to 

that for articles. 

Easy discovery and re-use hampered by 

fragmentation over journal silos. Not 

all supplements are linkable. 

Sometimes the files can only be 

accessed via the article and not 

independently. 

Volume is usually further restricted to 

underlying tables or explanatory 

graphs and full data sets are not 

included. 

Examples 

All peer reviewed scholarly journals 

Vast majority of STM journals and 

demand for this from authors is 

increasing rapidly lately. 

Some journals have made the 

availability of underlying research 

data (Nature, PLoS) a condition for 

publication. 

First examples appear of journals 

who can no longer handle the 

overload (Jnl of NeuroScience, Cell) 

Cell, Journal of Neuroscience, and a 

growing number of journals 

contemplating such restrictions as 

they find it hard to handle the 

growing volume and variety in 

formats (example: NISO/NFAIS 

working group) 



Community-endorsed Public 

Repository with bi-directional 

linking to and from articles 

3.a Journals have set up an own 

storage facility for data 

3.b Journals offer dynamic data 

made interactive, data can reside 

with the article or in public 

repositories 

Journals dedicated to so-called 

Data Publications only 

Data resides in a place where proper 

curation and formatting is secured, as well 

as discoverability and reuse. The 

bidirectional linking ensures connections 

with the publications. Author credits are 

indirect. 

Authors can be ensured that data is well 

curated and gets the right metadata 

attached for findability and check of 

provenance. Data remains closely 

connected to the article and becomes part 

of the public record of science. 

Graphs that show the underlying data via 

an extra click add depth to a research 

article. Data remain in the context of the 

article and become reusable at the same 

time. 

Data are described in-depth in these 

publications, facilitating findability, 

interpretability and re-use. Data remain 

in larger repositories and can be combined 

with other datasets. 

Data creators get the full credit of a public 

record of science. The data becomes 

citable. 

Life science area is well covered with 

many initiatives, in other areas first 

initiatives emerging. 

Many disciplines still lack a common 

solution. 

Future of existing repositories 

sometimes under threat because of 

pending cuts in government funding. 

Data will be spread over many different 

journals and may end up fragmented 

over silos, hampering reuse across 

different platforms. 

Usually only applied to data 

presentations in the article, not to 

(large) raw data set as such. 

Peer review of large sets of data is a 

challenge. 

Curation and preservation is in hands 

of the repository. 

The system requires persistent 

bidirectional linking to work well. 

Table 5: Categories to publish research data. 

Most journals in molecular biology 

and life sciences list these 

repositories and require authors to 

deposit there and submit the 

accession numbers of the databases. 

Strong supporters for this approach 

are PLoS, Science, Nature. 

Pangaea is an often cited example in 

earth sciences, collaborating with all 

Elsevier journals in this area. 

Thieme, for its chemistry journals in 

collaboration with FIZ and TIB 

The Biochemical Journal by Portland 

Press, experimental special issues 

by the Optical Society in America 

(OSA), 

Elsevier for data in the World 

ProteinDataBase (PDB) 

Earth Systems Science Data working 

with Pangaea. 



In the following paragraphs, each of these categories will be further explained and 

illustrated with real life cases. 

3.2. Common practice: Supplementary Material to Journal Articles 

The large majority of journals accept research data in supplementary files. From the 

results of the PARSE.Insight 29 survey we know that most journals accept research data 

in supplementary files: 

Can Authors submit their underlying research rch data with their publication? 

Graph 18 from PARSE.Insight 2 survey: N = 134 Publishers 

If we weigh in the size of the publishers (see PARSE.Insight report; the 3 % largest 

publishers publish 70 % of all journal articles and they all accept supplementary files), 

then over 90 % of all journals accept supplementary files with research data. 

Publishers generally accept a wide range of file and data formats (again: source 

PARSE.Insight): 

29 PARSE.Insight survey, see http://www.parse-insight.eu/downloads/PARSE-Insight_D3- 

4_SurveyReport_final_hq.pdf 



Graph19 from PARSE.Insight survey: What file formats does your journal accept ?, N=134 

Publishers 

As was shown in the introductory chapter (Chapter 1) of this report, we know that more 

than half of the researchers surveyed for PARSE.Insight would like to submit their data 

together with their manuscript to journals and publishers. This unveils a trend of likely 

growth in the submission of supplementary data files, because the present percentage 

who do so is below 20 % at the moment. Publishers confirm this growth trend as will 

become clear from several examples provided in this chapter. 

The instructions that these researchers find for most of the journals are fairly 

straightforward. This is the general instruction to authors from a large publisher with 

more than 2000 journals (ic Springer, bold text by us) 30 : 

accepts electronic supplementary material (animations, movies, 

audio, large original data, etc.) which will be published in the online version along 

with the article or a book chapter. This feature can add dimension to the article, 

as certain information cannot be printed or is more convenient in electronic form. 

30 Springer author instructions, see 

http://www.springer.com/authors/manuscript+guidelines?SGWID=0-40162-12-339499-0 



And from a similar size other publisher we find very similar instructions for most of 

their journals (bold typeface by us) 31 : 

Elsevier accepts electronic supplementary material to support and enhance your 

scientific research. Supplementary files offer the author additional possibilities to 

publish supporting applications, high-resolution images, background datasets, 

sound clips and more. Supplementary files supplied will be published online 

alongside the electronic version of your article in Elsevier Web products, 

including ScienceDirect. 

With a similar preference for a certain file formats as found with other publishers. 

Some journals have started to make the availability of underlying research data 

conditional for acceptance of the article by the journal. See for example the text by 

Nature in its authors instructions (underlining added by us) 32 : 

An inherent principle of publication is that others should be able to replicate and 

build upon the authors' published claims. Therefore, a condition of publication in 

a Nature journal is that authors are required to make materials, data a and 

associated protocols promptly available to readers without undue qualifications in 

material transfer agreements. Any restrictions on the availability of materials or 

information must be disclosed to the editors at the time of submission. 

Nature says that supporting data must be made available to editors and peer-reviewers 

at the time of submission for the purposes of evaluating the manuscript. But the journal 

does not say where authors should make their materials, data and protocols available for 

readers and users, except for a few exceptions regarding public repositories that will be 

mentioned in the next paragraph. The same is the case for the open access journal PLoS 

One 33 : 

PLoS is committed to ensuring the availability of data and materials that 

underpin any articles published in PLoS journals. We believe the ideal is that all 

data relevant to a given article and all readily replaceable materials be made 

immediately available without restrictions (whilst not compromising 

confidentiality in the context of human-subject research). (….) 

Failure to comply with this policy will be taken into account when publication 

decisions are made. We encourage researchers to contact journal editors if they 

encounter difficulties in obtaining data or materials from published articles. PLoS 

reserves the right to post corrections on articles, to contact authors’ institutions 

and funders, and in extreme cases to withdraw publication, if restrictions on 

data/materials access come to light after publication. 

How effective these mandates are is not known and anecdotal evidence points at a long 

way to go still. In an empirical study of data sharing by authors publishing in PLoS 

31 Elsevier author instructions, see an example at: 

http://www.elsevier.com/wps/find/journaldescription.cws_home/505772/authorinstructions#87000 

32 Nature instructions to authors on availability of data: 

http://www.nature.com/authors/policies/availability.html 

33 PLoS: http://www.plosone.org/static/policies.action#sharing 



journals, it is reported that requests for data underlying the published articles were only 

successful in one case out of ten, in spite of the journals’ clear policies (Savage and 

Vickers 2009) 34 . 

3.3. New limits on Supplemental Files to journal articles: restrictions to supplements 

The notion is growing that scholarly journals struggle to handle the exponential growth 

in supplementary data files and the responsibility for them in terms of securing 

permanent access and long term preservation. As an effect, the Journal of Neuroscience 

has declared a new policy in 2010 to no longer accept supplementary materials at all. 

Instead, authors may host supplemental material on an external web site and include in 

their article a footnote with a URL pointing to that site and a brief description of its 

contents. But reviewers and editors will no longer evaluate the supplemental material, 

the article as submitted will be treated as a self contained entity. 

At their journal site we can read the reasoning for this there is a clear indication that 

the exploding volume and the burden for peer review was becoming a real obstacle 

(underlining added by us) 35 : 

Although The Journal has published electronically since 1996, 

supplemental material first appeared around 2003. Since then, the amount of 

material associated with a typical article has grown dramatically (….) The sheer 

volume of supplemental material is adversely affecting peer review. 

The related editorial even speaks of a ‘proliferation among authors’ adding more and 

more material in the article supplements. 

In a similar context, we find a new policy by the journal Cell, implemented in October 

2009 and again referring to the ever growing amount of supplementary material 

(underlining added by us) 36 : 

Supplemental Information is a useful resource for presenting essential supporting 

materials online, and Cell Press is committed to the publication of these 

materials. However, as the amount of Supplemental Information has grown, it 

has become increasingly difficult for authors, reviewers, and readers to navigate 

due to the volume of information and the lack of defined structure and limits 

its. To 

address these problems, we are introducing these guidelines, which we believe 

will make Supplemental Information more useful and accessible to readers. 

The restrictions that Cell puts in place concern a limit on volume and total number of 

supplementary items: 

The total number of supplemental data items of all types (figures, tables, movies 

and other) per paper may not exceed two times the number of figures and tables 

34 Savage and Vickers 2009 in PLoSOne: 

http://www.plosone.org/article/info:doi/10.1371/journal.pone.0007078 

36 Cell editorial: http://www.cell.com/retrieve/pii/S0092867409011817 



in the main paper. For example, a paper with 7 main figures can have up to 14 

supplemental items total, of which up to 7 may be figures. 

Another restriction from Cell is that the supplemental material must bear a direct 

relationship to the main conclusions and content in the paper and must stay within the 

same scope: 

Supplemental Information is limited to data and other materials that directly 

support the main conclusions of a paper but are considered additional or 

secondary support for the main conclusions, or cannot be included in the main 

paper for reasons such as space or file format restrictions. Supplemental 

Information should be within the conceptual scope of the main paper and not 

extend beyond it. 

In a quote in The Scientists (February 2011) 37 , Emilie Marcus, Editor-in-Chief of Cell 

Press Journals, says "It had become a limitless bag of stuff." The publisher did not 

consider abolishing supplementary materials altogether because they have a diverse 

readership, with different levels of interest in a study's details, Marcus explained, but it 

was necessary to rein it in. "I do think there are different solutions for different 

journals," Marcus said. "Scientific communities and journals have probably not given 

enough thought to what to do with this capacity for supplemental materials. That needs 

to evolve." 

In 2010 a working group was set up by NISO/NFAIS 38 with a charge to define best 

practice recommendations on how publishers can best treat supplementary information, 

in terms of inclusion, handling, display and preservation. These guidelines are expected 

to appear in the second half of 2011. They will cover best practices for supplementary 

files to journals and also roles and responsibilities for availability, findability, quality 

control and preservation. 

3.4. How safe is data in supplementary journal article files? (or: quality and preservation 

ervation 

of supplementary journal article files) 

How do publishers treat the data in supplementary files and how safe is the data there? 

A complaint that publishers sometimes receive is about the lack of transparency over 

whether the supplementary files received from authors have been edited at all, peer 

reviewed or checked on format and general quality. Many publishers will just offer the 

author the service of posting the files in connection to the article exactly in the way they 

were received from the author. In some cases the files were part of the peer review 

process, in others they were supplied after acceptance of the article. Some publishers 

transfer the supplementary files into pdf’s before posting, which does not serve the reusability 

or further and deeper analysis of the data. 

The ways in which the supplementary files are provided with metadata vary widely 

between publishers. Some leave this to authors entirely which does not add to 

37 The Scientist: Supplemental or detrimental? - The Scientist - Magazine of the Life 

Sciences http://www.the-scientist.com/news/display/58027/#ixzz1NRVKAV6F) 

38 NISO, see http://www.niso.org/workrooms/supplemental 



consistency across data files. Also the mention of supplementary files in the article or the 

reference from the article to the supplements follows many different practices. 

Librarians find it sometimes a struggle to make sure that a literature search they carry 

out for research groups contains all related supplements (see Chapter 4). The 

NISO/NFAIS recommendations for Best Practice as mentioned earlier aim to advocate a 

clearer and a more common practice for this across publishers that help librarians find 

supplementary material when it exists. 

Again the findings from PARSE.Insight confirm this picture on many of its aspects. The 

137 publishers responding to the PARSE survey say about research data submitted by 

the author: 

• only 51 % of publishers validates the data submitted, mostly checking the file 

formats 

• only 44 % facilitates direct links to it 

• 39 % requires © transfer (against 57 % not) 

• 70 % has no preservation measures in place for the supplemental data other than 

for the articles 

As shown in the following survey results : 

Graph 20, source PARSE.Insight, N=137 publishers 







Do you have digital preservation arrangements for underlying research data 

Graph 23 PARSE.Insight survey: (N= 137 publishers) 

3.5. Data in community-endor 

endorsed public databases, linked to journal articles. 

In a number of subject areas, community archives have emerged where researchers are 

requested to deposit their data. Some of these examples were already mentioned in 

Chapter 2 of this report. Publishers have adopted these practices and a large number of 

journals encourage the authors to deposit their data there, rather than sending it along 

with their article. 

The advantages of this are manifold: the databases become more comprehensive, the 

data becomes better discoverable and can be used in combination with other data, and 

the connection with publications is ensured via bidirectional linking. 

PLoS One has been advocating such a policy quite extensively 31 : 

1. Data for which public repositories have been established and are in general use 

should be deposited before publication, and the appropriate accession numbers or 

digital object identifiers published with the paper. 

2. If an appropriate repository does not exist, data should be deposited as 

supporting information with the published paper. If this is not practical, data 

should be made freely available upon reasonable request. 

3. The conclusions of a study must not be dependent solely on the analysis of 

proprietary data. If proprietary data were used to reach a conclusion, and the 

authors are unwilling or unable to make these data public, then the paper must 

include an analysis of public data that validates the conclusions so that others 

can reproduce the analysis and build on the findings. 

PLoS also adds about the ideal of data sharing: 

We appreciate, however, that this ideal is not yet the norm in all fields. We are 

therefore currently collaborating with a number of subject-specific initiatives in 

order to develop relevant policies. In the meantime, authors must comply 



with current best practice in their discipline for the sharing of data via databases: 

for example, deposition of microarray data in ArrayExpress or GEO; deposition of 

gene sequences in GenBank or EMBL; and deposition of ecological data in 

DRYAD. We encourage all authors to comply with available field-specific 

standards for the preparation and recording of data. 

A similar policy encouraging authors to deposit their data in ‘approved databases’ is 

followed by the journal Science. Their instructions say 39 : 

Science supports the efforts of databases that aggregate published data for the 

use of the scientific community. Therefore, appropriate data sets (including 

microarray data, protein or DNA sequences, atomic coordinates or electron 

microscopy maps for macromolecular structures, and climate data) must be 

deposited in an approved database, and an accession number or a specific access 

address must be included in the published paper. 

For those cases where such an approved appropriate repository does not exist, Science 

wishes to have the datasets on its own website as supplementary material or at least 

hold the material in escrow if the files are hosted on a institutional website by the 

author. 

Further reading of the instructions seems to indicate that Science is also struggling with 

certain types of supplementary material in a careful balance to avoid having to absorb 

too much and still ensure that the underlying material can be examined by its readers. 

The journal asks authors to follow special procedures for making large or complex 

datasets available. See their special instructions for complex supporting data at their 

website. 

While these well known titles are rather specific in their support for community 

endorsed databases, it has become an established custom in many areas for publishers to 

collaborate with the main data archives. Typical author instructions for data deposits in 

public archives, linked to the publications, are (source Elsevier 29 ): 

If your article contains relevant unique identifiers or accession numbers 

(bioinformatics) linking to information on entities (genes, proteins, diseases, etc.) 

or structures deposited in public databases, then please indicate those entities 

according to the standard explained below. 

Authors should explicitly mention the database abbreviation (as mentioned 

below) together with the actual database number, bearing in mind that an error 

in a letter or number can result in a dead link in the online version of the article. 

Please use the following format: Database ID: xxxx 

And most publishers will specify the databases that allow links from (and increasingly 

to) the article: 

39 Science: see http://www.sciencemag.org/site/feature/contribinfo/prep/gen_info.xhtml#dataavail 

and http://www.sciencemag.org/site/feature/contribinfo/prep/prep_online.dtl and 

http://www.sciencemag.org/site/feature/contribinfo/prep/prep_online_special.xhtml 



Links can be provided in your online article to the following databases (examples 

of citations are given in parentheses): 

• GenBank: Genetic sequence database at the National Center for 

Biotechnical Information (NCBI) (GenBank ID: BA123456) 

• PDB: Worldwide Protein Data Bank (PDB ID: 1TUP) 

• CCDC: Cambridge Crystallographic Data Centre (CCDC ID: AI631510) 

• TAIR: The Arabidopsis Information Resource database (TAIR ID: 

AT1G01020) 

• NCT: ClinicalTrials.gov (NCT ID: NCT00222573) 

• OMIM: Online Mendelian Inheritance in Man (OMIM ID: 601240) 

• MINT: Molecular INTeractions database (MINT ID: 6166710) 

• MI: EMBL-EBI IntAct database for Molecular Interactions (MI ID: 0218) 

• UniProt: Universal Protein Resource Knowledgebase (UniProt ID: 

Q9H0H5) 

Findability, interpretability and re-usability are best served if the database ensures 

links back from the data(sets) to the articles about that data. Elsevier put this in place 

in 2010 via a collaboration with earth data archive PANGAEA. Datasets deposited at 

PANGAEA are automatically linked to corresponding articles in Elsevier journals on its 

electronic platform ScienceDirect and vice versa. A single click brings the user from the 

data to the ScienceDirect article, or reversely from ScienceDirect to the underlying data 

at PANGAEA, by means of DOI’s, both for the article and the dataset, (see 

Elsevier/PANGAEA press release (Elsevier 2010) 40 . 

Elsevier summarizes the process in 5 simple steps: 

1. Author submits article to publisher 

2. Author submits data set to repository 

3. At article publication, repository links article DOI to associated data set DOI, 

creating actual connection 

4. User sees link to ScienceDirect from PANGAEA 

5. User sees link to PANGAEA from ScienceDirect 

A few other databases and data-archives are also capable of providing links from the 

data to the corresponding articles, for example CCDC (Cambridge Crystallographic Data 

Centre) 41 and the PubChem Database 42 . These initiatives will make it much easier for 

authors to deposit the data in public archives as they can be ensured that future users of 

the data can easily find the corresponding articles. 

Important intermediary services have entered the field lately that help facilitate the 

workflow of authors and publishers for the parallel submission of data to a repository 

and the manuscript to the publisher, ensuring the bi-directional linking between 

publications and data in public repositories to be in place. DataCite and Dryad are very 

40 Elsevier press release: 

http://www.elsevier.com/wps/find/authored_newsitem.cws_home/companynews05_01616 

41 CCDC http://www.ccdc.cam.ac.uk/ 

42 PubChem http://pubchem.ncbi.nlm.nih.gov/ 



good examples further being explained in the next Chapter about Libraries and Data 

Centres. For most publishers, any system that uses DOI’s as persistent identifiers can be 

incorporated in the workflow. 

3.6. Data storage as a service by the journal 

We know of at least one example where the journal offers storage facilities for primary 

research data to the authors of their articles. In 2010 Thieme established, for its 

chemistry journals, an easy process by which the authors can submit primary chemistry 

data with their articles. Thieme, based in Stuttgart, works together with FIZ Karlsruhe 

(who also host their publications platform) and the TIB National Technical Library in 

Hannover. The process consists of 5 easy steps: 

1. At the same time with the article the author submits the research data to 

Thieme. 

2. Thieme hosts the research data in a data center (FIZ Karlsruhe). 

3. TIB assigns a DOI to the data. 

4. At the same time the article is published the primary data are published as 

independent entity but in connection with the article. 

5. The article quotes the research data as reference items with the assigned DOI. 

As their motivation for this new initiative, Thieme says in their press release 43 (link 

http://www.thieme.de/SID-4BE1BD47-99107897/connect-en/tc_oct_06_09.html ): 

In the field of chemistry, (…) data is accumulated by a variety of analytical, 

spectroscopic or computer simulation methods. Thus far, the vast amount of data 

lies scattered on the computers of scientists, who have produced the information. 

As no central repository exists, no archival storage is possible at the moment. 

Scientific results are solely published in journals – but not the primary data from 

which those results originate. Due to the missing credit that working up such 

data currently receives, primary data is often poorly documented, difficult to 

access and not saved for the long term. 

Dr. Susanne Haak, Managing Editor and responsible for the chemistry journals at 

Thieme explains, “Access to primary data is a fundamental condition for research work, 

particularly in the natural sciences.” Therefore, Thieme and experts from TIB have 

developed a uniform structure for publishing primary data. Through structuring and 

central data registration, a Germany-wide unique service of TIB, valuable knowledge 

will be harnessed. 

Since its inception at the end of 2009, the Chemistry journals of Thieme had 13 articles 

with data files added, per article roughly of the size of 5 – 10 MB. The data is not 

touched by Thieme, who simply collect them in ZIPfiles and check if the ordering of the 

subfiles is logical in the context of the article. The data files cannot be included in the 

43 Thieme press release http://www.thieme.de/SID-4BE1BD47-99107897/connecten/tc_oct_06_09.html 



(pdf) peer-review files and are not sent to the reviewers unless they request the Editor to 

see them. 

Thieme does not have the ambition to build up a database with chemistry information. 

The main aim is to provide individual datasets that support better understanding of the 

related research article. As Dr Susanne Haak explains: “The philosophy of a database is 

quite different from the collection of new, raw data. Everything that is entered into a 

searchable database needs to be carefully verified (PubChem is the best example) 

whereas raw data from experiments in chemistry may turn out to just help a chemist to 

understand what happened during a reaction.” She adds: “ CIF files that might be 

submitted to us as part of primary data have usually also been submitted to CCDC 

before as this is since decades THE database for crystallographic structure information”. 

3.7. Articles with interactive data 

In previous chapters we have explained the data problem as not just about availability 

and findability. It is also about interpretability and re-usability (see Chapter 1). Several 

publishers have undertaken initiatives to offer services around research data underlying 

the article that improve interpretability and sometimes even re-usability. 

The Biochemical Journal 44 , published by Portland Press offers dynamic pdf’s for its 

articles in their so-called Utopia Documents. These provide extra features that turn 

graphs into tables and tables into graphs from which the users can start viewing and 

using the data in their own fashion. These are available in excel sheets, so can be reused 

in other calculations. 

The Optical Society of America (OSA) together with the American NLM started a 

Interactive Science Publishing (ISP) project 45 in 2008 to enable authors to submit their 

data and figures to journals and to give editors and readers of their journals the 

possibility to view, analyze and interact with the source data connected with a scholarly 

article. Their main focus was the journal Optics Express. OSA supplies special software 

(from their ISP: Interactive Scholarly Publishing division 43 ) that help the reader to apply 

all visualization features for the underlying material to the article. ISP allows authors to 

publish large 2-D and 3-D datasets with original source data that can be viewed and 

analyzed interactively by readers. 

Elsevier provides in its articles a visualization and interaction applet for all related data 

that authors deposit in the PDB, the Worldwide Protein Databank. The app allows 

readers of the article to choose from several presentation formats to investigate the 

protein structures, in 2D or 3D, rotating or still. Elsevier emphasizes via this example 

that publishers need not be restrained in their offering of high-value added services for 

the analysis of data by the fact whether they store the data themselves. The data can 

just as easily be available in a public repository, available to all and available to applets 

to run on the data. In close collaboration with the NCBI, Elsevier offers a genome Viewer 

44 BioChemical Journal http://www.biochemj.org/bj/default.htm and 

http://www.biochemj.org/bj/424/bj4240317add.htm 

45 OSA-ISP project http://www.opticsinfobase.org/isp.cfm 



for all gene sequence data deposited by the author in GenBank. This viewer can also be 

applied from within the Elsevier article 46 . See Illustration 1: 

Illustration 1: Screen view of Gene-viewer on ScienceDirect, Elsevier 46 

3.8. Special Data Publications and Data Papers 

With the advent of ever growing volumes of datasets, and the urge for more sharing and 

open availability of data, there have been numerous suggestions for a new phenomenon 

of journals specially dedicated to so-called data publications. We examine in this context 

one of the very first initiatives in this space, namely by the Journal of Earth Systems 

Science Data 47 . 

This journal, as it states on its website: 

aims to establish a new subject of publication: to publish data according to the 

conventional fashion of publishing articles, applying the established principles of 

quality assessment through peer-review to datasets. The goals are to make 

datasets a reliable resource to build upon and to reward the authors by 

establishing priority and recognition through the impact of their articles. 

The journal sets as a very strict condition that the data described is deposited in a long 

term repository and lists several for which collaboration has been established. 

Other criteria for the data sets described are: 

• Persistent Identifier: The data sets have to have a unique and persistent 

identifiers, e.g. doi, ARK, etc. 

46 Example from the journal Genomics: http://dx.doi.org/10.1016/j.ygeno.2010.06.001 

47 ESSD: see http://www.earth-system-science-data.net/ 



• Open Access: The data sets have to be available free of charge and without any 

barriers except a usual registration to get a login free-of-charge. 

• Liberal Copyright 

ight: Anyone must be free to copy, distribute, transmit and adapt 

the data sets as long as he/she is giving credit to the original authors (equivalent 

to the Creative Commons Attribution License). 

• Long-term Availability: The repository has to meet the highest standards to 

guarantee a long-term availability of the data sets and a permanent access. 

Since its launch in 2010, the journal has accepted around 30 articles in three volumes, 

most of which Special Issues. The papers describe the data, the planning, 

instrumentation and execution of experiments or collection of data. Any interpretation of 

data is outside the scope of its regular articles. Articles on methods describe nontrivial 

statistical and other methods employed, e.g. to filter, normalize or convert raw data to 

primary, published data, as well as nontrivial instrumentation or operational methods. 

Any comparison to other methods is out of scope of regular articles. The peer review is 

public and follows an open discussion format on their website. 

The peer-review which checks on uniqueness, usefulness and completeness as well as 

quality, ensures that the data sets are: 

• at least plausible and contain no detectable problems; 

• of sufficiently high quality and their limitations are clearly stated; 

• open accessible (toll free), well annotated by standard metadata (e.g., ISO 19115) 

and available from a certified data center/repository; 

• customary with regard to their format(s) and/or access protocol, however not 

proprietary ones (e.g., Open Geospatial Consortium standards), expected to be 

useable for the foreseeable future. 

The main aim of the journal appears to promote (re-)usability of research data: 

The articles in this journal should enable the reviewer and the reader to review 

and use the data, respectively, with the least amount of effort. To this end, all 

necessary information should be presented through the article text and references 

in a concise manner and each article should publish as much data as possible. The 

aim is to minimize the overall workload of reviewers, e.g., by reviewing one 

instead of many articles, and to maximize the impact of each article. 

[…]It is clear that some of these quite abstract criteria may soon unfold to more 

(technically) specific ones, depending on the discipline or type of data. If 

necessary, the editors will try to make sure that more specific help for authors as 

well as for reviewers will be developed over time. (…) 

To help streamline the review process, a more formal list of criteria has been 

developed, which may serve as a checklist. 

In a personal interview conducted for this study, Hans Pfeiffenberger, one of the two 

Editors-in-Chief, explains how over 20 articles were part of special issues and so far less 

than 10 spontaneous submissions were received, mainly because authors are not yet 



accustomed to this new type of publication. From the peer review reports it is not 

completely clear how deep the reviewers have examined the data, but referees do tend to 

look at methodology and value of the data. The main task of the Editor is to check if the 

data are deposited in a safe repository and are persistently accessible. He thinks there is 

still an educational task to perform on the way peer review can help check and improve 

the quality of data. 

Pfeiffenberger also sees a need for better standards on the citation of data, especially 

which version of data to be cited: in its raw form, cleaned up or as gritted data, or only 

the final data product ready for re-use ? He quotes Oxford scholar David Shotton who 

advocates that data should be cited as: ‘first described in” and then link to a paper. 

The journal has an open peer review process, making the comments by the reviewers 

available for further transparency of the journal’s policies. 

As well as ESSD, which started in 2009, more initiatives in this area have now been 

launched. One of them is the journal GigaScience 48 published by BioMedCentral. The 

journal, which opened for submissions in the summer of 2011 and works together with 

the Beijing Genomics Institute BGI, aims according to its website: “ to revolutionize data 

dissemination, organization, understanding, and use. An online open-access open-data 

journal, we publish 'big-data' studies from the entire spectrum of life and biomedical 

sciences. To achieve our goals, the journal has a novel publication format: one that links 

standard manuscript publication with an extensive database that hosts all associated 

data and provides data analysis tools and cloud-computing resources”. 

It is likely that other journals will follow this example, possibly also in a hybrid way, 

including data-articles as a new article type for existing journals, in the way the 

International Journal of Robotics Research accepts Data Papers 49 : “Data papers are 

short (circa 4 pages) submissions that support and summarize a substantial archival 

data set which has itself been peer reviewed with the same diligence that regular 

submissions receive. The contribution is expected to be in the quality and utility of the 

data to the robotics community”. 

Similar discussions have already appeared on the blogs around PLoS. 

3.9. Gap analysis. 

In the previous paragraphs we have provided an overview of emerging practices for the 

integration of data and publications. Descending the Data Publication Pyramid, we find 

that data have always been part of the traditional literature in a much aggregated way. 

In the recent decade new extended forms of data presentation have found their way into 

online supplements to journal articles. Initially the absence of volume restrictions for 

added data was a blessing. Their recent proliferation has caused a halt and new 

48 GigaScience Journal http://www.gigasciencejournal.com/ 

49 Data Papers in IJRR 

http://www.uk.sagepub.com/journalsProdDesc.nav?prodId=Journal201324&crossRegion=eur#tabv 

iew=manuscriptSubmission 



limitations are being put in place to keep the data added to journal manuscripts relevant 

and manageable. 

At the same time, certain disciplines show the emergence of community endorsed data 

centres supported by scientific journals that include the archives’ accession numbers for 

links from the publication, or even add interactive viewers within the article to study the 

data in the archive within the context of the article. 

With more and more data deposited in archives, a new publication type has emerged, 

that of the Data Article. 

Returning to to the particular problem of making research data conform to the list of 4 

items of Chapter 1 in terms of: 

• Availability 

• Findability 

• Interpretability 

• Reusability 

we can make a rough rating for the journal practices described here above against these 

4 criteria (the greener the higher the overall rating is): 

Availability Findability Interpretability Re-usability 

+ +/- ++ - 

Data presented within 

articles 

Data in Journal 

Supplements, unrestricted ++ ++ +++ +/- 

Data in Journal 

Supplements, but 

restricted 

+ ++ +++ - 

Data in public archives, 

linked to publications +++ ++++ +++ +++ 

Journals storing data + ++ +++ ++ 

Journals making data 

++ +++ ++++ ++++ 

interactive 

Data Publications ++++ ++++ ++++ ++++ 

Table 6: Rating the different ways to publish research data against the Treloar criteria 

The most common current practice, which is to add data in supplementary files to 

journal articles, is clearly not the most ideal one if measured against these 4 criteria. 

Data deposited in public, community endorsed archives seems to have a better future, 

especially in terms of re-use and also for availability and findability. This is particularly 

true if they are accompanied by proper data publications that describe them and make 

them interpretable and re-usable. At the same time we know that only a few discipline 

areas have these public, community endorsed archives. Indeed, some of these archives 

appear to be threatened by cuts in government spending. 

If we add the three additional criteria of Chapter 1, namely citability, curation and 

preservation, the ratings or each of the current practices are: 



Citability Curation Preservation 

Data presented within articles +++ +++ +++ 

Data in Journal Supplements, 

unrestricted ++++ +/- +/- 

Data in Journal Supplements, but 

restricted ++ ++ + 

Data in public archives, linked to 

publications ++++ ++++ ++++ 

Journals storing data +++ +++ +++ 

Journals making data interactive ++++ +++ ++++ 

Data Publications ++++ ++++ ++++ 

Table 7: Rating the different ways to publish research data against the criteria of citability, 

curation and preservation. 

Curation and preservation of data in journal supplements is not always ensured by 

publishers as we see from the PARSE.Insight data and the specific requirements for 

specific formats of data, against an ever growing variety of data formats (including 

multimedia) asks for better solutions . Public repositories are better placed to deal with 

this issue. Bidirectional linking and data publications can ensure better citability of 

data. For the citability of data there are two aspects that play a role: the data as such 

should be citable (via DOI’s, accession numbers of other persistent identifiers) but in 

addition, the people who created or generated the data deserve credit. An increasing 

number of data repositories provide citation means for the data itself (including the links 

to and from the data), but very few conventions exist for the way the people behind the 

data get citation counts and credits for the work. The best way to do that currently is 

following the traditional way people are cited, via a publication that describes the data 

or has the data included. 

3.10. From raw data to processed data to data interpretations. 

In 2007 a majority of the larger publishers in the STM arena undersigned the so-called 

Brussels declaration which states 50 : 

Raw research data should be made freely available to all researchers. Publishers 

encourage the public posting of the raw data outputs of research. Sets or sub-sets 

of data that are submitted with a paper to a journal should wherever possible be 

made freely accessible to other scholars 

In this context, it is important to emphasize that the statement concerns raw data, not 

processed data, or data presentations or the data-interpretations as presented in a 

journal article. Using the Data Publication Pyramid as presented in Chapter 1, the 

Brussels statement concerns the public posting of raw data, the base layers of the 

pyramid. Preferably this public posting is done in an aggregated way in community 

endorsed archives for specific subject fields. It can be expected that many scientific 

disciplines will see a growing need to have such common solutions available that allow 

50 http://www.stm-assoc.org/brussels-declaration/ 



interlinking with journal publications. Project Dryad is one such example, the 

establishment of DataCite in 2009 another one (both described in Chapter 4). 

For the middle area in the pyramid, that of Data Selections, Processed Data and Data 

Representations, several options exist. These can be cared for in the context of journals 

or in the context of archives and databases, depending on the level of processing of the 

data. Traditionally these kinds of data have been included in supplementary files to 

journals and this custom is expected to grow further. Increasingly, we can expect more 

and more journal policies to raise the level of required data selection and processing 

(example: Cell) and no longer accept anything and everything in the supplemental 

materials but instead pose restrictions on volume and format along the criteria of 

relevancy and manageability. 

For the apex in this pyramid, data interpretations are included in a publication; there 

are no indications of paradigm shifts. But it is likely that data and publications will 

integrate further, at different levels and in more novel ways. There will be new 

innovations in the way the data are displayed and presented that confirm certain 

analyses and conclusions. Examples given in this chapter concern data made interactive 

from within the article via protein viewers and genome viewers. 

3.11. Diverging and Converging Trends. 

We see diverging trends taking place as well as converging trends in the way publishers 

are handling the increasing amount of data alongside articles. This is probably a strong 

indication that this area is in transition at the moment. 

A clear example of diverging trends can be found in: 

• Journals have more and more data submitted in supplementary files, and most of 

the journals accommodate this, also for a growing variety of file formats. At the 

same time, the first few examples now exist where journals could no longer 

handle this flow in view of the sheer volume and hence have stopped accepting 

supplementary files or have put limitations for them. 

Whereas a converging trend is emerging in these areas: 

• More journals support the principles of data sharing and data availability and 

press (or even mandate) authors to deposit data in public archives and to follow 

the conventions for this in their subject area. 

• More publishers collaborate with community endorsed, public archives to make 

data and publications inter-linkable and citable, thereby endorsing the Brussels 

declaration in practice, and with positive effects on the integration of data and 

publications, their findability, discoverability, interpretability and re-use. 

• A growing number of publishers offer services to present data in more 

sophisticated and even interactive ways, that increase their interpretability and 

hence their re-use further. 



3.12. Opportunities for publishers in Data Exchange 

From the practices and laudable initiatives gathered and analyzed in this research 

study, we can summarize the following elements as important opportunities for 

publishers to further improve the integration of data and publications. 

• Require availability of underlying research material as an editorial policy 

(example: Nature, PLoS) 

• More careful treatment of digital research data submitted to journals and ensure 

it is stored, curated and preserved in trustworthy places (several examples of 

collaboration with community endorsed repositories) 

• Ensure (bi-directional) links and persistent identifiers (examples for listed public 

archives, DataCite, Dryad) 

• Establish uniform citation practices (examples Elsevier-PANGAEA, ESSD, 

DataCite, Dryad, Thieme) 

• Establish common practice for peer review of data (example ESSD) 

• Develop data-publications and quality standards (example ESSD, GigaScience, IJ 

Robotics Research) 

In order to offset these points against the listed issues around data we create the 

following table: 


Availability 

Findability 


Re-usability 

Citability 

Curation 

Preservation 

Publishers opportunity to help improve situation: 

Articles with data provide richer content and higher usage 

Impose stricter editorial policies about availability of underlying data 

which is in line with general funder’s trends 

Ensure data is stored in a safe place, preferably a public repository 

Be transparent about curation and preservation of submitted data 

Ensure bi-directional links between data and publications 



Provide services around data such as viewer apps for underlying data 

within the article or interactive graphs, tables and images 


Interactive data from within articles 

Links to the relevant datasets, not just the database 


Establish uniform data citation standards 




Transparency about curation of submitted data 


Transparency about preservation of submitted data 


Table 2 repeated: Data Opportunities for Publishers. 



In general, we can say that publishers will tend to follow their authors’ wishes. With the 

trend clearly towards researchers who share more and more data, funders who make 

this conditional, and libraries and archives working towards better accessibility and 

retrievability of data, publishers can play an important role in the integration of data 

and publications for the sake of better discoverability, interpretability and re-sue of 

research data. 



4. DATA CENTRE AND LIBRARY PERSPECTIVE 

This chapter describes how libraries and data centres respond to the increasing amount 

of data that is produced and available and how they support availability, findability, 

interpretability, and re-usability of data. As such we assume that libraries and data 

centres deal, or have to deal, with any level of data that researchers and/or publishers 

want to make available: selected data representations, data collections and structured 

databases, raw data and original data sets. Based on desk research, we describe the 

current practice and rationale for action in libraries and data centres. We elaborate on 

the implications of increasing data integration in publication workflows, and present 

exemplary data initiatives, in which libraries and data centres are involved. Contact 

persons of each initiative were addressed with key questions, and their responses 

informed the analysis. The chapter ends by highlighting gaps as well as opportunities. 

4.1. Libraries and data centres as a 

custodians of data 

Libraries and data centres are traditionally positioned at opposite ends of the research 

lifecycle: Data centres help researchers collect and process their data, and libraries deal 

with the publications that result from research projects. Libraries also help arrange the 

input at the start a new research cycle: the search for publications as the basis for new 

research. With the convergence of data and publications, and interdependencies between 

data and journal publications, such traditional roles become blurred. Reports like 

“Riding the Wave” 51 recognise that the requirements of e-science and enhanced scientific 

publishing necessitate a comprehensive infrastructure for scientific information. 

Libraries and data centres have important, partly overlapping, but mostly 

complementary roles to fulfil. 

To create an infrastructure that systematically supports such data publication scenarios, 

libraries and data centres must align, or create new common conventions in data 

description and identification, and balance the relation between disciplinary 

particularities and large-scale interoperability. In this process, libraries and data centres 

complement each other: 

Research data centres can be best considered in this context as the experts in their 

respective research disciplines and in handling their discipline specific data. They are 

set up to support data creation and access. They provide research teams with storage 

space, and services around data creation as well as preservation, and they provide 

academics and other users with access to data files and with training and advice on how 

to use them. They are familiar with data protection and privacy regulations and with 

research ethic issues. They are well positioned to adopt new and emerging types of data 

in their discipline and increasingly sophisticated methods of record-linking and 

statistical matching. They are aware of data quality issues and existing disciplinary 

standards. 

51 E.g., Riding the Wave: How Europe Can Gain From The Rising Tide of Scientific Data. Final 

report of the High Level Expert Group on Scientific Data (2010). 

http://ec.europa.eu/information_society/newsroom/cf/document.cfm?action=display&doc_id=707 



Libraries have been keepers of knowledge for hundreds and thousands (in the case of 

Library of Alexandria) of years. They are experts in categorizing fields of knowledge and 

in recording and cataloguing all relevant information about a particular publication, 

including provenance and information about its author. They have been collecting, 

organizing, describing, preserving and making available knowledge and information 

manifested in printed books and articles and they are ready to transfer this experience 

to new forms of (digital) collections. 

Several symposia and publications bear witness that the libraries community is in a 

transition process, rethinking their role in an increasingly digital environment in 

general and in e-research in particular: 

A study commissioned by the Research Information Network (RIN) in 2007 asked if data 

management was a job for academic librarians. The survey data provided mixed 

messages: “Many librarians see data curation as a natural extension of their current 

role, but there is also evidence of caution in terms of the curation of large-scale datasets 

linked to e-research.” 52 

A symposium organized by the US Council on Library and Information Resources (CLIR) 

explored functions of the research library in a changing information landscape in 2008. 53 

It came to the conclusion that libraries needed to engage in data management and data 

curation to reflect basic changes in how scholars work – in collaboration with faculty and 

publishers. Rick Luce suggested in the same publication that traditional library roles 

must be augmented by new capabilities, centred on collaborative, data-intensive 

information resources. 54 

The Association of European Research Libraries, LIBER, made “Scholarly 

Communication” one of its 5 strategic priorities 2009-2012. 55 The corresponding working 

group on e-science recently organised the workshop “Libraries and research data: 

exploring alternatives for services and partnerships” that met large interest (presently 

not published). 

While the library system has evolved over the centuries, the data centre landscape is a 

relatively young one, with the first scientific data centres dating back to the mid 20 th 

century (ex: US National Climatic Data Centre: 1951, World Data Centre: 1957/8). The 

data centre landscape is more fragmented than the library system. There are wellestablished 

disciplinary data centres in some data intensive research domains (see for 

example CESSDA member organisations for the Social Sciences, CERN computer centre 

for Particle Physics, the members of the International Virtual Observatory Alliance in 

52 RIN 2007: Researchers’ Use of Academic Libraries and their Services. A report commissioned 

by the Research Information Network and the Consortium of Research Libraries (2007). 

http://www.rin.ac.uk/system/files/attachments/Researchers-libraries-services-report.pdf 

53 CLIR 2008: No Brief Candle: Reconceiving Research Libraries for the 21st Century. Council on 

Library and Information Resources (2008). http://www.clir.org/pubs/abstract/pub142abst.html 

54 Richard E. Luce: A New Value Equation Challenge: The Emergence of eResearch and Roles for 

Research Libraries. In: No Brief Candle: Reconceiving Research Libraries for the 21 st Century 

(2008). http://www.clir.org/pubs/reports/pub142/pub142.pdf 

55 http://www.libereurope.eu/committee/scholarly-communication 



Astronomy or the said World Data Centres for geophysical data), but with few exceptions 

barely visible centres in the Humanities. 

One noteworthy example of a publicly funded, national disciplinary data archive, with a 

vast collection of digital data in the Social Sciences and Humanities is the UK Data 

Archive (UKDA). It was established in 1967 by the UK Social Science Research Council, 

which has so far provided the long-term commitment of funds. 

The Life Science community is characterized by numerous specialized data centres, e.g., 

GenBank at the US National Center for Biotechnical Information, the Worldwide 

Protein data Bank, Cambridge Crystallographic Data Centre and many more. 

4.2. Common practice and rationale for action 

Libraries and data centres serve the needs of the research community and in that role, 

must react to increasingly demanding user needs (see chapter 2 on researchers’ 

perspective) and support increasingly sophisticated and complex publishers’ products 

(see chapter 3 on publishers’ perspective). In a society where information is available 

abundantly and often for free on the internet, libraries and data centres are under 

pressure to strengthen their role as professional information suppliers. 

Another influencing factor is institutional requirements: Libraries and data centres are 

increasingly confronted with data management requirements from their funding bodies, 

involving them in data creation during the research workflow. Many academic institutes 

in a growing number of countries 56 have adopted Open Access policies which require 

their data centres to provide publishing support to the university’s research groups. 

More and more funders oblige their grant recipients to make their data (openly) 

available after the end of a project (see also “Who pays for what” in chapter 2). There is a 

natural expectation that libraries and data centres will support the principal 

investigators with data management plans and to provide secure storage space for the 

created data. 

A large evidence base on the common practice as well as on trends in the scientific 

infrastructure is available from the PARSE.Insight project. As far as libraries and data 

centres are concerned, however, the underlying figures must be interpreted carefully. 

The PARSE.Insight report never consistently defined the difference between data 

centres and libraries as we try to do for this report. The stakeholder group “data 

management” of the PARSE.Insight survey was composed of 7 archives, 20 data centres, 

152 research libraries, 13 regional institutes, 24 national libraries and 3 institutions 

that identified themselves as “other” (see Graph 24). 

56 See ROARMAP, the Registry of Open Access Repositories Mandatory Archiving Policies: 

http://roarmap.eprints.org/ 



Graph 24, Source PARSE.Insight, N=241, background of ‘Data Management’ respondents 

The questionnaire distributed in the stakeholder group “data management” is available 

online, as are Graphs of most survey results. 57 Here, a special selection of the findings 

most relevant to data sharing are reanalyzed. The participating libraries and data 

centres agreed predominantly that data preservation was important or very important 

for the following reasons: 

• Publicly funded research output should be properly preserved (98%) 

• Preserved data stimulates the advancement of science (96%) 

• It allows for re-analysis of existing data (95%) 

“Researchers want someone else to pay for data preservation!” 

The PARSE.Insight survey showed that researchers look for an organisational structure 

to invest in data curation (see Graph 25, also chapter 2). In agreement with the high 

awareness for the importance of data preservation, especially libraries consider 

themselves responsible to fulfil this role: 

57 http://www.parse-insight.eu/downloads/PARSE-insight_survey_questions_datamanagement.pdf 

and https://www.swivel.com/people/1015959-PARSE-insight/group_assets/public 



Graph 25, Source PARSE.Insight, N=241, ‘Data Management’ respondents, N = 77 

However, in practice, only 44% of the responding institutions accept research data for 

storage and preservation (Graph 26). 




It is likely that the percentage is unevenly distributed between the group of 

participating data centres and that of the libraries. If 100% of the participating data 

centres and perhaps some or most of the archives accept research data for storage and 

preservation and only a very small proportion of the libraries actually accept research 

data, this result can either indicate a large gap on the side of the libraries or indicate a 

strategy of specialization and division of labour. 

It is likely that the percentage is unevenly distributed between the group of 

participating data centres and that of the libraries. If 100% of the participating data 

centres and perhaps some or most of the archives accept research data for storage and 

preservation and only a very small proportion of the libraries actually accept research 

data, this result can either indicate a large gap on the side of the libraries or indicate a 

strategy of specialization and division of labour. 

“Researchers want to be in control of their data!” 

According to the findings of PARSE.Insight (see chapter 2) and to findings of the SURF 

foundation 58 , it is of paramount importance for authors that they keep control of their 

data: “In all cases, when the data is transferred to another party, researchers wish to 

remain in control of their data.” 59 Consequently, libraries and data centres, as keepers 

of authors’ data must make sure that they respect this wish. 

58 SURF 2010: Martin Feijen: What researchers want. A review of literature describing what 

researchers want with regard to storage of and access to research data. Commissioned by the 

SURF Foundation (2010). 

http://www.surffoundation.nl/en/publicaties/Pages/Whatresearcherswant.aspx 

59 SURF 2010 



71.5% of the PARSE.Insight data management survey participants stated that they had 

security protocols in place that protect stored data from unauthorized modification, 

damage or deletion. In 19.2% of the participating organisations, action remains to be 

taken (Graph 27). 


However, only 54,1% confirmed that they had procedures to determine ownership and 

for identifying and managing data rights – an important criteria for researchers to 

entrust an organisation with their data (Graph 28). 


“Researchers rs want credit for sharing their data!” 

Another key topic for researchers is that if they make their data available, it should be 

visible and it should be possible to receive credit for it. 




At present, only 54% of the libraries and data centres support linking to stored data from 

journal articles (Graph 29). However, efforts are under way to facilitate exactly this. 

DataCite, for example, undertakes efforts to make research data citable and accessible in 

an internationally harmonized way 28 . 

The statistical findings suggest already that there is a high awareness in libraries and 

data centres, but not yet a comprehensive preparedness to take on the challenge. 

4.3. Implications of data integration for libraries and data centres 

There are good reasons for fostering the integration of research data and publications. 

As shown in the introduction of this report (see Chapter 1) integration of publications 

and research data has the potential to facilitate findability and re-usability of data, and 

to provide authors with better credits for their data. It also adds value and background 

to the publications. 

Publishers offer several ways in which data and publications can be integrated already 

(see chapter 3). Data centres are, to a high degree, part of new publishing models, 

supporting data creation in the first place, or when publishers require authors to deposit 

underlying research data in public data archives and link to it. After all, many 

manifestations of data, as illustrated by the Data Publication Pyramid in the 

introductory chapter of this report (Graph 1) have an impact on libraries and data 

centres: 

1. Data contained and explained within the article 

Implication for libraries/data centres: Prepare for adequate preservation 

strategies. The preservation of enriched articles may be more demanding than 

preservation of traditional articles. Novel ways of embedding data within the 

article (clickable graphs that provide underlying tables) will require more 

sophisticated preservation means. 



2. Data published in supplementary files to articles 

Implication for libraries/data centres: Ensure that article and supplementary files 

stay together and that presentation and preservation mechanisms for 

supplementary files are in place. 

3. Datasets referenced from the articles and held in data centres and repositories 

Implication for libraries/data centres: Distributed responsibility between article 

holder (publisher, library) and data holder (data centre or publisher). Datasets 

must be citable. The link between article and referenced data must be persistent. 

Presentation and preservation mechanisms must be ensured. If the data resides 

in a publisher‘s storage facility, perpetual access and eventually hand-off 

mechanisms to either libraries or data centres must be developed. 

4. Data published independently from written publications, e.g. in databases or 

special data journals (“data publication”) 

Implication for libraries/data centres: Support publication processes. Datasets 

need curation and special treatment that considers the granularity and dynamics 

of research data. Add metadata to datasets for documentation, to support re-use, 

and to facilitate search and retrieval of data. 

5. Data in drawers and on disks at the institute 

Implication for libraries/data centres: Support scientists in preparing data 

management plans at an early stage of the research process to avoid “isolated 

data holdings” in the first place. Develop user friendly, low-threshold data 

publication or data deposit services. 

To summarise, libraries and data centres must support data publishing as a prerequisite 

for data availability, including persistent identification/citation of datasets, and 

solutions for data description, documentation and retrieval, which together facilitate 

findability. They must also ensure long-term data archiving including data curation and 

preservation as a condition for data interpretability and re-usability. Libraries and data 

centres have started to enter into new alliances (as will be described in more detail in 

the next section) to develop new strategies together or with other actors such as 

publishers or research institutions. They can be involved on several levels, e.g. as active 

service operator, as provider of a specific sub-service (e.g. assignment of persistent 

identifiers), or as custodian of the results of such services. 

4.4. Libraries and data centres engagement in new services and alliances 

Durign the course of our research we found several new flagship projects and initiatives 

where libraries and data centres are probing the integration of data and publications on 

different levels. We present examples for persistent identifier and linking initiatives 

(findability – DataCite), data publication and data management support (Availability & 

interpretability – Dryad, Dataverse), and data archiving (re-usability – Pangaea). 

In DataCite, libraries and data centres have allied to establish easier access to scientific 

research data online. 60 The goal of the international consortium is to make research data 

60 http://datacite.org/index.html 



citable and accessible in a harmonized, interoperable and persistent way. 15 members 

from 10 countries – among them 6 Data centres or data services and 9 libraries – create 

and maintain an infrastructure to register data sets and assign unique persistent 

identifiers to them. 

By using DOI names for data registration DataCite offers a simple and well known 

solution for citing data from publications. Users – researchers, publishers, libraries – can 

use the same technical infrastructure for datasets that they already use for research 

articles. 

The focus of DataCite is the registration of data sets and assigning of persistent 

identifiers, not on storage and preservation of research data. The responsibility for the 

research data, including access, remains with the data centres or other trusted 

institution. The content holders are held responsible for quality assurance, metadata 

creation, storage and access. DataCite however provides supports, e.g. the DataCite 

Metadata Scheme 61 , created by a working group of DataCite members. Also, a DataCite 

working group is defining criteria for trustworthy data centres with the rationale that 

stable descriptions of the duties and technical requirements for data centres which are 

using DOI names are needed. 62 With such criteria, DataCite would create an instrument 

that helps publishers and researchers trust that research data is stored in a reliable and 

persistent way. 

One of the greatest achievements of DataCite is the inclusion of all relevant players in 

this arena: data centres, libraries, and publishers. DataCite acts on the assumption that 

progress in integrating data and publication can only be achieved by a joint effort of 

these stakeholders, combining strengths and influence of each of them. As the DataCite 

initiators confirm to us, one of their main short-term goals is to raise awareness in the 

editorial boards to allow referencing of datasets in publications. In the DataCite concept, 

libraries are not expected to act as data centres themselves, but continue to be a source 

of information for researchers. They should open their catalogues to scientific data and 

other content types and mediate access to data in data centres as remote content. 

DataCite addresses data from science and technology alike and in principle spans all 

disciplines. The German Leibniz Institute for the Social Sciences (GESIS) runs a pilot 

DOI registration agency for social science research data based on the DataCite 

infrastructure. 

DryadError! Bookmark not defined. is an example for an international repository that is 

committed to the “long tail” of data from the more decentralised biosciences, where data 

is not necessarily kept in large-scale repositories, or from under-financed fields such as 

the humanities: 

Dryad is designed to preserve the underlying data reported in a paper at the time of 

publication, when there is the greatest incentive and the ability for authors to share 

their data. This is particularly important in the case of data for which a specialized 

61 http://datacite.org/schema/DataCite-MetadataKernel_v2.0.pdf 

62 Klump 2011: Jens Klump: Criteria for the Trustworthiness of Data Centres. D-Lib Magazine, 

Volume 17, Number 1/2 (2011). http://www.dlib.org/dlib/january11/klump/01klump.html 



repository does not exist. Datasets are assigned with persistent identifiers to enable data 

citations. 

Dryad was developed in the US by the National Evolutionary Synthesis Center and the 

University of North Carolina Metadata Research Center, in coordination with journals 

and societies in evolutionary biology and ecology. The initiative resulted from a 

workshop dedicated to “Data Preservation, Sharing, and Discovery: Challenges for Small 

Science in the Digital Era” in May 2007. To journals and societies in these “smaller” 

sciences, Dryad offers a shared solution for data publication and archiving, thus 

relieving them from developing solutions on their own which would only lead to a 

fragmented landscape. 

In the UK, there are attempts to integrate Dryad as a building block in the national 

scientific infrastructure. The initiative “Dryad UK” 63 is a 12 month JISC funded project 

and run by The British Library and Oxford University with in partnership with several 

associate organisations. The project aims at moving Dray to a sustainable business 

proposition and establishing a UK mirror of the Dryad repository. In order to reach this 

goal, a business models are being developed and publisher expansion is being 

promoted 64 . Cost recovery is a realistic and achievable goal but establishing the most 

appropriate model for this is the challenge of Dryad. For example, a authors may fund 

Drayd submissions from their research grants. 

The Dryad UK initiators confirm to us that they have begun discussing potential 

business models together with publishers and funders. They point out the beginning 

integration of the Dryad repository with large publishers, e.g. PLoS and BiomedCentral 

for the first time, and new publisher workflow integration, allowing for peer review of 

the data behind an academic publication. They also acknowledge that initial resistance 

from some large commercial publishers, who are wary of imposing a system on all of 

their journals, or who still plan to explore commercial opportunities in the field 

themselves, could be overcome. 

While DataCite and Dryad deal with readily produced datasets at the moment of 

publication, Dataverse is an example for an initiative offering data management support 

throughout the research life cycle, thereby preventing that data gets lost in disks and 

drawers in the first place: 

The Dataverse Network 65 is an application to publish, share, reference, extract and 

analyze research data. It started as collaboration between the Harvard-MIT Data Center 

(now Institute for Quantitative Social Science) and the Harvard University Library and 

is presently implemented in Social Science disciplines. However, Dataverse collaborates 

with researchers and archives to expand the Dataverse Network as a data management 

and publishing framework beyond social science. 

63 http://datadryad.org/dryaduk 

64 See Beagrie et al: Business Models and Cost Estimation: Dryad Repository Case Study. 

Proceedings of iPres 2010. http://www.ifs.tuwien.ac.at/dp/ipres2010/papers/beagrie-37.pdf 

65 http://thedata.org/home 



The open source Dataverse Network software allows for implementation of individual 

virtual archives, so called “dataverses”. An institution can implement several of such 

dataverses and in this way distribute ownership between multiple researchers or 

research groups. That way, Dataverse addresses the desire of scientists to maintain 

control of their data, in particular to keep sovereignty over restrictions for their data sets 

(important for interpretablility and for re-use). At the same time, Dataverse offers a 

central repository infrastructure with support for professional archiving services such as 

back ups, recovery, and persistent identification. 

A dataverse is designed to contain data organized in studies. Each study comprises the 

actual data files, complementary files, and metadata. Persistent identifiers are allocated 

to studies and can be used in publications to point to the respective evidence base. By 

allowing owners of the data to create persistent identifiers and a citation for their 

datasets even before public release of the data set, Dataverse accommodates the fear of 

authors that their data may be (mis-)interpreted by others before their own analyses are 

published. The owner of the Dataverse data can still use the persistent reference in an 

article, and then release the dataset once the article is published. 

The Dataverse representatives point the data citation mechanism out as a major 

achievement of their initiative. The Dataverse Network software generates 

automatically a data citation for each data set published in a dataverse. Before the 

project, citations of data were inconsistent or nonexistent in many publications, which 

made data retrieval highly uncertain. Dataverse facilitates referencing data sets from 

publications in a standardized and persistent way. 

Barriers are seen in lacking knowledge among researchers of the services that allow 

them to share data, in insufficient recognition and incentives to researchers for 

publishing their data sets, and in inconsistency between requirements from funding 

agencies or journals on publishing data, and providing insufficient funding to continue 

implementing and maintaining framework solutions. 

Pangaea can be considered a representative initiative in the area of disciplinary data 

publishing and curation: 

Pangaea 66 is an information service run by the German Alfred Wegener Institute for 

Polar and Marine Research (AWI) and the Center for Marine Environmental Sciences 

(MARUM), also located in Germany. It focuses on archiving, publishing and distributing 

georeferenced data from earth system research. 

Various international research projects use Pangaea as their data repository. Thereby, 

Pangaea addresses a need of publicly funded projects, which are often required to store 

primary data for a defined period after the end of the project. The German regulation, for 

example, sets forth that “Primary data as the basis for publications shall be securely 

stored for ten years in a durable form in the institution of their origin.” 67 Because not all 

66 http://www.pangaea.de/ 

67 Recommendations of the Commission on Professional Self Regulation in Science: Proposals for 

Safeguarding Good Scientific Practice (1998) 



institutions can provide adequate storage capacity, let alone an infrastructure 

supporting description, persistent identification, search and retrieval of data, this task 

can be fulfilled by disciplinary data services. 

Pangaea has developed and offers a range of archiving and publishing services. For 

example, it acts as publishing service and long term archive for the World Data Center 

for Marine Environmental Sciences. Pangaea aims at firmly establishing the concept of 

data publishing in the Marine Environmental Community within the next couple of 

years. Linking data to journal articles is an important part in this process and it 

happens in a bi-directional way: articles link to the data in Pangaea, and the Pangaea 

data point and link to the articles that use the data. Pangaea is already a designated 

archive for the Earth Science journals of Elsevier (see chapter 4) and is also the ‘home’ of 

the new data journal Earth System Science Data (ESSD). 

The people responsible for Pangaea consider it the most important achievement of 

Pangaea that it has established collaborations with science publishers for cross 

referencing of science data and articles. DataCite DOIs are used for persistent 

identification of datasets. The Pangaea data publication process follows established 

publication processes with submission including a metadescription, formatting rules, 

abstract, archiving with lead time for proof-read, defining a citation, registration and 

final publication. Search and retrieval of the data sets via library catalogues is ensured 

through cooperation with the German National Library of Science and Technology 

4.5. Gaps and dilemmas 

The findings of the PARSE.Insight project, the initiatives presented in this chapter, and 

the sheer number of workshops, conferences and publications related to data 

management, data sharing, (open) access to data suggest that the need for action has 

been recognised in the library and data centres communities alike. They are actively 

involved in developing persistent identifier systems for research data, data citation 

standards, and solutions for data description, documentation and retrieval, as well as in 

data curation and preservation. 

Availability 

In terms of available infrastructure, plenty of solutions and possibilities are already 

available for the often mentioned problem of making research data available. There are 

vast possibilities for researchers to make their data available via institutional or 

disciplinary repositories, and increasingly together with publications. A challenge is that 

not all researchers are aware of the services available to them. Another challenge may 

be that the multitude of possibilities may create a fragmented landscape. Here, 

especially research libraries as information suppliers have an important role to play: 

They should engage with researchers to raise awareness for good data management 

http://www.dfg.de/download/pdf/dfg_im_profil/reden_stellungnahmen/download/self_regulation_9 

8.pdf 



practices, the benefits of data sharing, and the options available in the different 

disciplines, but also acceptance for the best and most reliable services available for a 

specific discipline. As Martin Feijen highlights in a report by SURF 2010: “Researchers 

must be in control of what happens to their data, who has access to it, and under what 

conditions.” 68 

When addressing researchers as both data users and data creators, a division of roles 

and responsibilities seems suitable between libraries and data centres with libraries 

addressing researchers in general, of all professional levels and all kinds of disciplines, 

and data centres advising the professional users. On the service level, it seems suitable 

that libraries focus on data retrieval, cataloguing, registering (“retrieval focus”), and 

data centres focus on ensuring the long term availability of published data, including 

data storage, backups and replication in multiple locations (“data management focus”). 

Findability 

A precondition for proper data retrieval is that good metadescriptions are added to the 

datasets and that links that lead to datasets either from metadescriptions or from 

publications are persistent. Metadata schemes must reflect the granularity of research 

data, the relationship between data sets, and the often frequent updating of datasets. 

With persistent identifier initiatives like DataCite or the Data Document Initiative in 

the Social Sciences 69 , some research communities are on a good track for common 

solutions. The challenge lies in aligning progress in all scientific disciplines – not 

necessarily between disciplines, but at least within disciplines. Retrieval services only 

make sense if researchers and institutions within an institution adhere to certain 

common conventions. Most data centres provide data retrieval services for their own 

data base. Libraries that integrate data material in their catalogues to facilitate 

findability and access are still a rare exception. One of the rare examples is the “GetInfo” 

service of the German National Library of Science and Technology, where different 

internal and external databases can be searched at one time (Illustration 2). 

68 SURF 2010 

69 http://www.ddialliance.org/ 



Illustration 2: Screen view of the GetInfo service by TIB, Hannover, Germany. 


To interpret research data created by others, good descriptions and documentation must 

be available (as for findability). The more documentation available, the easier it is for 

researchers to interpret other researchers’ data. Documentation can range from 



descriptions of the data to so-called data publications all the way to (linking) the full 

publication using the data. Services like Pangaea that require researchers to submit 

metadescriptions with their data and adhere to certain formatting conventions (so that 

all datasets can be interpreted in a similar way) are a solid beginning. Crosslinks 

between articles and data are another means to support interpretability, because 

verbalized interpretation of the dataset in a publication helps the understanding of the 

original dataset. While links from articles to data become increasingly common, the 

other way around from data to articles is not yet so widely used, but good examples exist: 

e.g., Pangaea, PubChem and the Cambridge Crystallographic Database Centre. From a 

technical viewpoint, the interpretability of datasets can be ensured by separating them 

from vulnerable data carriers like CD-ROMs or DVDs and storing them on hard drives, 

including backups, forward migration and replications. Data centres seem to be best 

equipped to take on this challenge. In disciplines where there are no established data 

centres (yet), the universities institutional data centre, well equipped libraries, or library 

federations or initiatives like Dryad UK should stand in, although this may perpetuate 

the risk of fragmentation. 

Re-usability 

Ensuring re-usability is the most difficult goal of data management in a data centre and 

library setting. In addition to all the preconditions needed to ensure interpretability, reusability 

often requires software to be available for analysing the datasets. The 

researcher who wants to re-use another researcher’s dataset does not only need 

intellectual, discipline specific understanding of the available datasets, but also the 

skills to operate the appropriate software. Besides constant monitoring of the data 

holdings, libraries and data centres need to maintain format and software registries to 

plan for data preservation actions. First approaches to preservation of scientific data 

were for example, developed in the CASPAR project 70 , and are followed up in the 

APARSEN network of excellence 71 , but continued research is needed. 

General dilemmas 

Altogether, the many new initiatives in the area of data integration are promising. 

However, against the expected explosion of research data (see chapter 1 and 2) they are 

still more or less exceptional cases. There are a couple of pioneering libraries, often 

embedded in big and capable universities and involved in several initiatives at one time. 

The danger is that a few actors master the transition to a data-intensive scholarly 

information infrastructure well, and that the majority of stakeholders follow in a passive 

manner. 

70 http://www.casparpreserves.eu/ 

71 http://www.alliancepermanentaccess.org/current-projects/aparsen 



The results of the PARSE.Insight Gap Analysis of the “Scientific Libraries” community 

confirm this danger. 72 Although it was focused on preservation of digital research data, 

its tendency can easily be transferred to the wider area of data integration. 

Overall, the PARSE.Insight Gap Analysis indicated a gap between better and less 

prepared libraries. This gap could be found in almost all analyzed areas: 

The more data a library has to store, the better it considers itself prepared and 

responsible for digital preservation – to the point that funding for digital preservation 

will be an important issue for the libraries. Another relation was shown between the 

amount of data stored at a library and the implementation of data and access 

management strategies. Libraries with preservation and selection policies in place have 

smaller preservation gaps (in terms of preservation strategies implemented) than those 

who have not. Effectively, the PARSE.Insight Gap Analysis found those libraries which 

were slower in addressing the problem, were far behind the “early starters” in most 

categories. 

Another, similar gap becomes apparent between disciplines: Influential, policy setting 

data centres do not exist in all scientific communities. A step in the right direction are 

two large programs implemented by the European Commission, CLARIN in the in the 

Linguistics and Humanities, and DARIAH in the Arts and Humanities. 

Substantial and sustained funding is required to develop and market new services but 

often there is a tendency to address novel challenges in time-limited projects. While it is 

absolutely necessary to initiate a first step into action that way, it puts at the same time 

even promising results need to progress to sustainable services and this needs to be 

directed by libraries and data centres in partnership with their funding bodies 73 . 

4.6. Opportunities for libraries and data centres 

The issue of data availability and data re-usability gets a lot of attention from users, 

funders, and decision makers 51,74, . For libraries and data centres, this opens up the 

possibility to re-position themselves as complementary professional information 

providers in this field. In order to enable data availability, findability, interpretability 

and re-useability, libraries and data centres’ data managers need to be involved from the 

very beginning of the research process in order to ensure high data quality. 75 

Datasets differ in many important ways from publications (e.g. granularity, iteration 

rate, data can be dynamic) and libraries must adjust to these new requirements as part 

of their new role. The immediate future for libraries will likely be characterised by 

72 PARSE 2010: PARSE.Insight. Deliverable D4.3. Gap Analysis Final Report (2010). 

http://www.parse-insight.eu/downloads/PARSE-Insight_D4-3_GapAnalysisFinalReport.pdf 

73 Sustainable Economics for a Digital Planet: Ensuring Long-Term Access to Digital Information. 

Final Report of the Blue Ribbon Task Force on Sustainable Digital Preservation and Access 

(2010). 

74 Prepublication data sharing”, “Post-publication sharing of data and tools”, Nature: Vol 461/10 

September 2009 

75 Richard E. Luce (2008) 



emerging and new roles in librarianship professions. Considering the initiatives 

presented in 4.4, it seems obvious that there is a need for “data librarians”. The 

implementation of the “Datasets Programme” from The British Library 76 illustrates 

action in this area. Libraries in the US like the University of North Carolina at Chapel 

Hill Library or Penn State University Library have already developed data management 

toolkits to support researchers at the proposal stage. 77 

In all their actions, libraries and data centres as institutions serve the research 

communities. Many studies point out that researchers prefer local and discipline specific 

data management support and want to retain control of the data until research is 

published. 78 , 79 This balanced with the need to make data easily available and searchable 

suggesting there may be a role for libraries to act as an intermediary between 

researchers and larger data centres and in disciplines where there are no large data 

centres available, between researchers and the institutional repository. Overall, dialogue 

and interaction between the stakeholders is crucial and platforms to systematically 

enable it are desirable. 

76 http://www.bl.uk/datasets 

77 http://www.lib.unc.edu/reference/data_services/researchdatatoolkit/index.html, 

http://www.libraries.psu.edu/psul/scholar/datamanagement.html 

78 SURF 2010 

79 PARSE.Insight. Deliverable D3.6. Insight Report (2010). http://www.parseinsight.eu/downloads/PARSE-Insight_D3-6_InsightReport.pdf 




Availability 

Findability 


Re-usability 

Citability 

Libraries opportunity to help 

improve situation: 

Accept datasets for storage at 

library and/or open up library 

catalogue to research data sets to 

allow access to data at least as 

remote content. 

Data centres opportunity to help 

improve situation: 

Be open and transparent and lower 

barriers to researchers to make their 

data available via discipline or 

institutional data centre. 

Support of persistent identifiers. 

Engage in developing common metadescription schemas and common 

citation practices. 

Promote use of common standards and tools among researchers 

Provide metadescriptions to 

datasets. 

Support crosslinks between 

publications and datasets. 

Be transparent about conditions 

under which the data sets can be 

re-used (expert knowledge needed, 

software needed). 

Help researchers understand 

metadescriptions of datasets. 

Establish and maintain knowledge 

base about data and their context. 

Curate and preserve datasets. 

Archive software needed for reanalysis 

of data. 

Engage in establishing uniform data citation standards. 

Support and promote persistent identifiers. 

Curation/Preserv 

ation 

Transparency about curation of 

submitted data. 

Collaboration with data creators 

and data centres. 

Promote good data management 

practice. 

Transparency about curation of 

submitted data. 

Collaboration with data creators and 

libraries. 

Instruct researchers on discipline 

specific best practices in data 

creation (preservation formats, 

documentation of experiment,…) 

Table 3 repeated: Data Opportunities for Libraries and Data Centers. 



5. REPORT EPILOGUE: MAPPING PING THE ROAD AHEAD 

5.1. Can Libraries and Data centres fill the missing link? 

This report has shown that the key to ensuring the long term success of integrating data 

and publications is to ensure that the data are managed and preserved in such as way 

that they remain: 

• Available 

• Findable 

• Interpretable 

• Reusable 

• Citable 

These criteria are meaningful to, and act as incentives for, both researchers and 

publishers to engage in linking data to publications but neither group believe they can 

fulfil this role on their own. No one group has responsibility across the whole 

communication chain or has the resources to satisfy all of these criteria. So a key 

question we wish to raise here is; can librarians and data centres help fill this missing 

link ? To what extent do they have the existing relationships with researchers and/ or 

the related knowledge (including skills) to ensure that many of these criteria are met? 

We have established that our three stakeholder groups of researchers, publishers and 

libraries and data centres have much to gain from embracing the integration of data and 

publications. There are several opportunities to be grasped for researchers, publishers, 

and libraries and data centres. It is clear from the researcher’s perspective that making 

data available needs to be incentivised. Funding is one incentive, but it is important that 

researchers are credited for their data and that making data available will increase the 

author’s visibility, in other words the data should be easily citable. 

It is also clear that, whilst publishers, in principle, are open to further integration of 

data and publications, there are challenges associated with ensuring the quality and 

longevity of the data submitted. Accepting and storing data submitted in supplementary 

files can be hugely demanding on publishers’ resources. There is also the question over 

what level of data, as illustrated in the ‘Gray’s Pyramid, see Graph 1, a publication 

should accept and in what manifestation. 

There are exciting developments in publishing in relation to data, such as datapublications 

whose main aim is to describe available datasets. Publishers are investing 

in developing services to enrich publications with data and are doing so in collaboration 

with public archive services such as libraries and data centres. 

Data centres play an important role in the long term storage of data and it can be 

surmised that libraries have a supporting role in this landscape, whether that be in 

supporting researchers in storing their data or ensuring that data remains available to 

and is discoverable by the end user when starting up new research projects. It is clear 

that, for libraries, a priority is to ensure that quality data can always be accessed easily 

by their users. 



What is unclear is a definition of the roles that libraries should fill and what the 

incentives and barriers exist for researchers to work with libraries on data management. 

This lack of clarity calls for a dialogue to take place about why and how libraries should 

play a role in the integration of data and publications. There are gaps that need to be 

addressed as the preceding chapter identifies. By drawing on data from previous 

research and on real life examples it presents some arguments and outlines 

opportunities but a more complete picture may be drawn by engaging library 

professionals and researchers themselves in this dialogue. 

5.2. What does the Data Publication Pyramid mean for roles and responsibilities ? 

The Data Publication Pyramid has appeared throughout this report (see Graph 1). Every 

layer of the Data Publication Pyramid presents different challenges and calls for a 

variety of approaches to improve data exchange. As we descend the layers of the data 

pyramid the division of roles and responsibilities becomes less clear. 

The top layer, publications with data, is well established and already an integral part of 

the record of science and the systems around it for discoverability, access and retrieval 

are well in place. The associated roles and responsibilities between researchers, libraries 

and publishers are well defined and delineated. But it only presents a tip of the iceberg 

of available data. 

At this stage of the pyramid, there is limited potential for the reuse of data as the data is 

usually embedded in the article in an aggregated form. While this potential for re-use 

increases in the layer below that: Processed Data and Data representations, often 

presented in supplementary files to journal articles, the criteria of discoverability and 

longevity are less sure there. As a result, an increasing number of publishers encourage 

authors to submit their data to the third layer, that of Data Collections and Structured 

Databases, from where and to which the publications can link. 

At this level, the Data Collections and Structured Databases layer, boundaries begin to 

blur and roles and responsibilities of libraries and data centres need to be better defined. 

This layer also offers the most potential for data exchange as, at this level, the data are 

at their most findable, reusable, and well curated and preserved. 

The ‘long tail’ of data at the bottom of the pyramid, where data tends to remain in 

drawers and on disks of the institute, has been wholly the domain and responsibility of 

the researcher. That this data meets none of our criteria for data exchange is both a 

problem and an opportunity. As discussed in chapters 2 and 4, there are several reasons 

why this data has not been integrated. Lack of incentives, lack of local or discipline 

specific repositories, and fears over losing control over or credit for the data all 

contribute to this data remaining locked in their silos. Collaborative projects such as 

DataCite, Dryad and Dataverse are addressing these issues but more needs to be done to 

encourage researchers to start embracing data management and to make their data 

available. 



5.3. In dialogue with Research Librarians: LIBER Workshop 2011 

As research libraries are one of the key stakeholders in this report, ODE ran a workshop 

at the LIBER 2011 annual conference drawing on a draft of this report. Five speakers 

representing various backgrounds, including publishing, data centres, libraries, and 

research, were shown a copy of the draft report and asked to prepare a provocative 

statement related to their reading of the report. 

Speakers and Provocations 

“We need proper data citations in the reference section.“ Merce Crosas 

Merce Crosas, Project Director of the Dataverse Network at Harvard, reminded the 

audience that, for the most part, we all agree that making data available is important, 

that we have important reasons for it (advancement of science, verifyability of research 

results etc.), and that nowadays, we have the technologies to realize it. So, she asked, 

shouldn’t it really be only a small problem that we can easily solve? In her opinion, 

upgrading data citations in scholarly articles from in-text citations to full citations in the 

reference sections would be a big step forward in the right direction. 

“Libraries must tackle the long tail data problem.“ Brian Hole 

Brian Hole from DryadUK at The British Library stated that a major barrier to data 

exchange is the “long tail of data”, which means that a large proportion of research data 

sits on researchers computers and doesn’t get into a repository. These data are not made 

available for reuse and are almost inevitably lost. This is particularly true of Humanities 

data and Hole is convinced that there is a massive potential for re-use of these data. 

Hole argued against the prescription that smaller libraries should leave data curation to 

large data centres. When no specific subject repositories for the long tail data, which is 

particularly true of the humanities, it simply isn’t exploited, maintained or preserved. In 

Hole’s opinion, (research) libraries are very well placed to bridge this gap because they 

are already placed within the researchers’ workflow. At the very least they can educate 

researchers and provide data management plans. Beyond that, they may create and 

maintain their own repositories or act as advocates for the establishment of data 

repositories within their institutions. 



“Clear and easy to understand citation metrics for datasets and automatics 

mechanisms to count them are urgently needed“ Maurits Van der Graaf 

Maurits Van der Graaf, author of a SURF study on the quality of research data 80 , 

advocates establishing a Dataset Impact Factor as an incentive for researchers to 

publish and properly cite datasets, and a Data Archive Impact Factor as a means to 

further professionalize data archive management. The Data Archive Impact Factor could 

help data archives to measure their relevance to scientific research, and help funders to 

evaluate the effectiveness of their support. Van der Graaf expects that this would 

influence publishers in developing additional services linking to high impact data 

archives. 

“No Publication without Data – no data without publication” Eefke Smit 

Eefke Smit, Director of Standards and Technology at the International Association of 

STM Publishers, referred to the Data Publication Pyramid of this report to illustrate 

that currently only a very small fraction of all data created gets ever published. 

Smit argues for a change in practices because publication has a large potential to make 

data visible and re-usable. And in return data publication enhances the traditional 

publication by providing supporting evidence and background to the official Record of 

Science. Also, via citations it serves as a credit systems for the people behind the data. 

Moreover, Smit calls for constructive collaboration throughout the information chain, 

where data is not necessarily kept at the publisher, but rather securely stored and 

preserved in certified, reliable data repositories, and via persistent identifiers remain 

linked to publications – bidirectionally. 

“The research community needs to establish a common format for data d 

acquisition and 

interpretation” Rick Luce 

Rick Luce, Vice Provost and Director of Libraries for Emory University, drew the 

audience’s attention to the complexity of roles and responsibilities in the data landscape. 

No institution can soundly manage data over time when it doesn’t know what is allowed 

to do with it or how to treat the data correctly. Therefore, it is essential to clarify the 

question of ownership at the instant of data handovers. 

Thereby, three principles should be observed: 

• The data integrity principle: Ensure the integrity of research data to implement 

trust in research processes, enable researchers to verify published research 

results. 

80 http://www.dlib.org/dlib/january11/waaijers/01waaijers.html and 

http://www.surffoundation.nl/nl/publicaties/Documents/SURFshare_Over_kwaliteit_van_onderzo 

eksdata_dec2010DEF.pdf 



• The data access and sharing principle: Make data that is integral to publicly 

reported results publicly available. 

• The data stewardship principle: Provide proper data documentation, curation, 

and long term preservation to enable re-use. 

5.4. Emerging Issues 

Feedback from the workshop was recorded and five main issues or concerns emerged: 

• Citation Metrics 

• Roles 

• Why libraries? 

• Approaches to data publishing 

• Incentives 

Much of what came out of the workshop reflected the speakers’ provocations and 

reaffirmed the relevance of the issues that have been highlighted throughout this report. 

The following are some the issues that were emphasised by participants throughout the 

session. Due to the nature of the workshop, some opinions are in conflict and reflect the 

lively debate in the workshop. 

Citation Metrics 

Measuring the impact of published data may be central to the success of data 

publications and measuring the impact of publications is a preoccupation of research 

libraries. 

This is an area which requires further analysis. Journal Impact Factors, although 

established, have drawbacks; it may not be ideal to connect the impact of data to the 

impact of a journal, further it is important that whatever metrics are developed, they are 

simple and easy to interpret. 

Roles 

Unsurprisingly for a workshop with library professionals, the definition of roles was a 

major concern and drew many comments from the audience. Comments ranged from 

what roles Libraries should play in data management and exchange, to what new skill 

set library professionals need to meet the new challenges implied in these changing 

roles. 

It was acknowledged that libraries need to align themselves with what is a dynamic 

research life cycle, becoming more project-oriented rather than providing services on an 

as needed basis. This reflects Rick Luce’s (2008) assertion that libraries need to 

reposition themselves to become involved in research at the beginning of the research 

life cycle. 



The majority of participants believed that libraries should have a role as repositories of 

data. Some traditional library skills are applicable to data management, such as 

collection management and information retrieval skills and the libraries established 

relationship with researchers as teachers of information literacy skills means that they 

are well placed to provide guidance to researchers in the creation of data management 

plans. 

If libraries are to become more involved in the integration of data and publications then 

there may well be a necessity to develop software development skills to support the 

exploitation of data as well as expertise in the area of intellectual property (IP) in order 

to ensure that the rights of the creators of the data are protected. 

The increase in the use of integrated data means that boundaries between libraries, data 

centres and publishing are blurring, possibly, as one participant stated, becoming one 

intellectual unit, and this has perhaps the greatest implication for future roles. 

Why Libraries? 

As a further provocation participants were asked why libraries should have a role in 

data exchange? This question not only served to draw out the rational for the 

repositioning of libraries but also helped highlight some of the barriers to library 

involvement in data exchange. 

Reflective of an issue raised in chapter 2, researchers are not interested in archiving or 

curating data themselves and there is a real danger of fragmentation if these activities 

are left up to individual researchers. Libraries already have a track record in 

supporting researchers in their work. 

This could present an argument for institutional repositories but on the other hand, 

there may be a danger of fragmentation if libraries become repositories of data where 

discipline specific international resources such as PANGAEA or GenBank are available 

and a preferable solution for researchers. Also, libraries’ existing strengths lie in 

creating structure rather than storage. 

Approaches to data publishing 

Much of the dialogue in the workshop focused on the approach that should be taken to 

encourage, sustain and support data publishing and data integration. First of all, there 

was a general sentiment that there is a need for a definition of metadata for research 

data. 

There is also a need to focus on interoperability rather than specific subject domains, 

regions or institutions. This could pose a problem for libraries as existing library 

information infrastructures are not always easy to make interoperable. 

At the same time it is important to work with researchers and research groups at local 

and project level. As one participant put it “local effort can result in global solutions”. 



Incentives 

Incentives for researchers to publish their data were discussed in chapter 2. From the 

research library perspective it is important that there is an incentive for researchers to 

work with libraries in making this data available. 

Without a doubt citation and impact is a major incentive for researchers (hence the 

importance of citation metrics). In order for data to be widely cited it must be findable, 

reusable and citable. Librarians and libraries have the skills and resources to make this 

achievable. 

Another incentive that libraries may provide (if they act as repositories) is the promise of 

sustainability; the fact that an institution is taking care of the data, preserving it and 

curating it. This is not just a boon to researchers but a valuable addition to research 

proposals. 

5.5. The Next Step: Survey to Document Current and Project Future Roles 

What is clear from the workshop analysis is that there are many questions that need to 

be answered. Research Libraries are keen to engage in a dialogue about data exchange 

and there exists awareness that there is a need to reposition library institutions in this 

changing landscape. The workshop also served to validate the issues raised in this 

report. 

The report highlights an opportunity for collaboration. With the increasing demand from 

funders for researchers to make their data publicly available and the ensuing need for 

support in data management, and publishers supporting the principle of data sharing by 

signing up to the Brussels Declaration, libraries are faced with the opportunity to 

reposition themselves to become embedded in the research process. 

In order to mine the potential of the bottom layer of the data pyramid it is important to 

understand how researchers can be supported in making data more available than is 

presently the case. 

This report has provided evidence of the impact that data sharing and reuse has and can 

have on the scholarly communication chain and how important the integration of data 

and publications is. 

The survey among research libraries and researchers within their institutes that follows 

this report during the fall of 2011, aims to clarify further the roles of stakeholders 

concerned by measuring their awareness and readiness for more responsibility in 

research data. It draws from the findings of this report and from what we have learned 

from the LIBER workshop. It aims to reveal how stakeholders’ roles are changing. 

Perhaps more importantly, it presents information and guidance for the likely evolution 

of these roles -- to ensure the ongoing integrity of the scholarly record and for the 

creation of incentives for stakeholders to support this. 

Better insight in existing strengths versus weaknesses and opportunities versus threats 

should help create the conditions to increase data publication activity and, ultimately, 

data sharing and reuse.

Integration of Data and Publications - Alliance for Permanent Access

You also want an ePaper? Increase the reach of your titles

Delete template?

Save as template?